Tug-vimu: a GoPro-based mobility assessment bridging the gap between technology and clinical practice Authors: Sara Houde, Camille Martin, Karina Lebel Published: 2025 ABSTRACT BackgroundAccurate mobility assessment is critical for identifying individuals at risk of falls, particularly among older adults. While gold-standard motion capture systems offer high precision, their clinical adoption is often hindered by high costs, spatial constraints, and technical complexity.MethodsThis study introduces and validates TUG-VIMU, a low-cost, portable system for instrumented Timed Up and Go (TUG) testing that integrates a GoPro camera with an ArUco marker. The system fuses inertial data from the camera’s embedded sensors with pose estimation derived from ArUco tracking and video processing algorithms. A key innovation lies in the use of a velocity-adapted continuous wavelet transform for automatic, robust, and adaptive step segmentation. TUG-VIMU was evaluated in a cohort of healthy younger and older adults (n = 16 with 8 over 60, age = 45 ± 19 years), across a range of gait speeds.ResultsAutomatic trial and phase segmentation achieved a mean error below 0.37 s. Step event detection reached sub-50 ms accuracy, enabling reliable extraction of spatiotemporal parameters. Gait velocity at slow and preferred walking speeds was estimated with a mean error of 0.00 ± 0.02 m/s, and step length accuracy was within - 0.69 ± 1.97 cm. The combined inertial and video-based approach also enabled robust step detection during turning phases, an often overlooked challenge in gait analysis.ConclusionTUG-VIMU demonstrated high temporal and spatial accuracy, robust gait phase detection, and reliable estimation of clinically relevant parameters. Its performance was comparable to established motion capture systems, particularly at slower walking speeds, while offering enhanced accessibility, portability, and ease of use. These findings support the potential of TUG-VIMU as a practical and scalable tool for gait assessment in clinical and community settings. Future work includes automated reporting features, open-source distribution, and validation in populations with mobility impairments such as Parkinson’s disease. Introduction The Timed Up and Go (TUG) test is widely used in clinical settings to assess functional mobility and fall risk [1, 2]. It requires minimal equipment, a chair, a stopwatch, and offers a quick, global assessment. However, its reliance on total completion time and subjective observation limits both its specificity and its sensitivity to subtle changes over time. Indeed, the TUG test can identify individuals at risk of falling, but it does not provide insight into the underlying causes of mobility deficits. In contrast, analyzing spatio-temporal parameters (STP) during the different phases of the TUG test offers a more detailed understanding of movement impairments. Falls are often caused by improper weight transfer or body shifting during movement [3]. Specific STP can help assess these risks. For example, excessive trunk flexion during the sit-to-stand phase is shown to impair balance recovery and stabilization [4]; prolonged duration, reduced angular velocity, and increased number of steps and step variability during turning may reflect balance deficits or compensatory strategies [3, 5]; wider stance, shorter steps and slower gait speed are associated with instability and increased risk of mobility disability[6, 7]. However, such detailed assessments are difficult to perform using the traditional TUG setup. In laboratory settings, optoelectronic motion capture (Mocap) systems provide precise STP measurements using reflective markers positioned on anatomical landmarks and multiple infrared cameras. These systems are considered the gold standard for kinematic analysis due to their high accuracy and reliability [8, 9]. For instance, Caron-Laramée et al. [10] used a Mocap system to analyze gait in 25 healthy individuals and 21 living with psoriatic arthritic, achieving sub-millisecond accuracy in heel strike detection and 0.3 cm precision in stride length, compared to force plates [10]. Despite their precision, these Mocap systems are expensive and complex to use, limiting their clinical applicability[11, 12]. To address these limitations, Inertial Measurement Units (IMUs) have been explored as a more accessible alternative. IMUs, which include 3-axis accelerometers and gyroscopes, enable 3D orientation tracking. They are commonly used to detect gait events via ankle acceleration [13], and to identify turning or sit-to-stand transitions using trunk-mounted or lower back positioned gyroscope, respectively [14]. Although promising, IMUs face challenges related to calibration, environmental interference, and algorithm robustness [11, 12, 15–17]. Over the last two decades, clinical research systems based on IMUs have emerged, such as the Clario Opal V2R (formerly APDM Mobility Lab) and the Movella Xsens Awinda, along with their associated software platforms [18, 19]. These systems offer detailed motion analysis capabilities and have demonstrated strong potential for enhancing mobility assessment in research settings. However, their cost and, in some cases, the complexity of setup and data interpretation may hinder widespread adoption in routine clinical practice, particularly in resource-limited environments. More recently, video-based motion tracking using AI algorithms has gained attention due to its affordability and minimal hardware requirements. Applications in fields such as virtual reality, gaming, sports, and robotics have demonstrated their potential [20, 21]. While promising, clinical adoption remains limited due to variability introduced by lighting, clothing, and body morphology [21]. Among video-based methods, fiducial markers such as ArUco and AprilTags offer robust tracking capabilities and have been successfully applied in domains like augmented reality and robotic navigation [22]. To bridge the gap between technological potential and clinical applications, it is essential to evaluate the strengths and limitations of different approaches. In this context, combining IMUs with video (VIMU) data represents a promising strategy for enhancing movement analysis during the TUG test. This approach is supported by scientific literature, which highlights the benefits of integrating multiple data modalities. For instance, Yu et al. [23] developed a system for respiratory monitoring and posture analysis in surgical settings using on ArUco markers. Similarly, Basiratzadeh et al. [24] proposed a real-time application that estimates posture and range of motion using a smartphone and AprilTag technology. Notably, commercially available solutions like the GoPro camera, which integrates high-definition video with inertial sensors, offer significant potential for developing such assessment systems. This study introduces and validates the TUG-VIMU system, an instrumented version of the TUG test using a GoPro camera, enabling accessible and cost-effective objective mobility assessment. The specific objectives are to: Present the TUG-VIMU system, including its physical setup and data processing algorithms.Validate the segmentation algorithms against a laboratory-based gold standard kinematic assessment.Evaluate the STP parameters derived from the TUG-VIMU system in comparison to those obtained from a gold standard. Present the TUG-VIMU system, including its physical setup and data processing algorithms. Present the TUG-VIMU system, including its physical setup and data processing algorithms. Validate the segmentation algorithms against a laboratory-based gold standard kinematic assessment. Validate the segmentation algorithms against a laboratory-based gold standard kinematic assessment. Evaluate the STP parameters derived from the TUG-VIMU system in comparison to those obtained from a gold standard. Evaluate the STP parameters derived from the TUG-VIMU system in comparison to those obtained from a gold standard. TUG-VIMU system The TUG-VIMU system utilizes a GoPro Hero 9 camera (software version v02.50.00) mounted on the individual’s trunk using the GoPro Performance Chest Mount. The camera is configured to record videos in linear mode at a resolution of 4K and a frame rate of 60 Hz, ensuring a higher level of detail for the amount of motion captured. To ensure consistency and minimize processing artifacts, the following features are disabled: low light mode, video clipping, and hypersmooth stabilization. In addition to video capture, the GoPro Hero 9 is equipped with a built-in 3-axis accelerometer and gyroscope, which record raw inertial data at a sampling rate of 200 Hz. Before data collection, the camera’s field of view is calibrated using a LEGO-built checkerboard and the OpenCV camera calibration procedure to ensure accurate spatial measurements [25, 26]. The system also makes use of two ArUco markers (4x4 format, 96 mm x 96 mm) built with LEGO bricks to facilitate usability. The ArUco markers are placed at the extremities of the normal TUG environment, aligned with the participant’s sternum. Specifically, initial tests were conducted positioning the ArUco markers 1 m behind the chair and the turning point. These fiducial markers will be used during signal processing to allow distance information to be processed. The following subsections address data processing steps, including segmentation and STP derivation. Figure 1B presents an overview of the global workflow for signal processing.Fig. 1A TUG-VIMU system installation. B Data processing diagram representing the flow of data, starting from the acquisition to the analysis parameters A TUG-VIMU system installation. B Data processing diagram representing the flow of data, starting from the acquisition to the analysis parameters Signal acquisition and preprocessing The TUG-VIMU system enables acquisition of inertial data from the GoPro camera, which includes 3D acceleration, 3D angular velocity, and 3D orientation, all expressed in a camera body frame and collected at 200 Hz. The camera video feed is processed using Python OpenCV library Version 4.5.3 to extract the translation vector and rotation vector of the ArUco markers. All signals are then uploaded and processed in MATLAB R2023a (MathWorks, USA). Gravity is removed from the acceleration using the reconstructed gravity information. It is computed from the camera’s orientation raw data, obtained from GoPro’s software, and the initial stable moment. Relative orientation of the trunk is computed for further use. Anatomical alignment Anatomical alignment refers to the process of expressing the information collected in a reference frame aligned with the body. In this specific case, this is performed based on the gravity axis and the stand-to-walk motion. The ArUco translation vector is aligned from the camera’s point of view using the orientation quaternion raw data from the GoPro. TUG-VIMU segmentation The segmentation process takes place at different levels. Level one consists of isolating the beginning and end of the test. This is performed using a sliding window technique on the norms of the acceleration and gyroscope signals, searching for stable moments. End of stability, therefore, corresponds to test initiation, while the return to stability corresponds to the end of the test. A window of 0.5 s with 50% overlap was determined to be a good trade-off between precision and processing time [27]. The second level of segmentation consists of identifying and isolating the different phases within the TUG, namely the sit-to-stand/stand-to-sit transfer (STS), the gait segments, and the turn. This segmentation enables the targeted analysis of STPs. STS segmentation A single segmentation algorithm is used to detect both transfer phases, sit-to-stand and stand-to-sit, with the distinction between the two based on their temporal order: the first identified STS event is assigned to the sit-to-stand phase, and the last to the stand-to-sit phase. This algorithm is inspired by the work of Weiss et al. [28], who observed that transfer movements exhibit a characteristic sinusoidal (’S’-shaped) pattern in the angular velocity signal during the trunk flexion and extension (see Fig. 2). To extract this pattern, the signal is first processed using a discrete wavelet transform (DWT) with a sym4 mother wavelet across eight decomposition levels, focusing on levels seven and eight. A synthetic ’S’ wave, corresponding to the dominant frequency of the flexion-extension movement as determined by Fast Fourier Transform (FFT) analysis, is then generated. Each STS phase is identified by cross-correlating the angular velocity signal with this synthetic waveform (Fig. 2). The onset of the sit-to-stand phase is further refined using a threshold-based detection method adapted from Lepetit [29], using acceleration and gyroscope data. The threshold value is defined from the sensor’s noise at rest, using the rules of three sigma to represent 99.7% of the noise. Then the phase is considered complete at the first local minimum below zero following the peak amplitude of the STS event (Fig. 2).Fig. 2STS segmentation, localized from a cross-correlation with ’S’ wave signal STS segmentation, localized from a cross-correlation with ’S’ wave signal Gait segmentation and step events detection Gait segmentation is performed at two hierarchical levels. At the first level, gait segments are defined as the periods between the transfer and turning phases, resulting in two gait segments per TUG trial. Within each segment, a second level of segmentation is applied to identify individual gait events related to steps. Figure 3 illustrates the relationship between gait, gait cycles, steps, and key gait events, including heel strike (HS) and toe-off (TO). Accurate identification of these events is essential for the computation of spatio-temporal gait parameters.Fig. 3The gait cycle parameters overview The gait cycle parameters overview Figure 4A presents an overview of the gait processing workflow for TUG-VIMU, highlighting the dependencies between segmentation and STP computation. The segmentation algorithm is based on the Continuous Wavelet Transform (CWT), following the approach proposed by McCamley et al. [30] and Del Din et al. [31]. This method enables visualization of body motion as a pendulum-like oscillation, either mediolateral or anteroposterior, with each footstep (see green line in Fig. 4Bii). In this context, the term footstep refers to the occurrence of a full gait cycle, not solely the HS event. To improve robustness across varying walking speeds, the CWT is adapted to the participant’s gait frequency, as recommended by Caramia et al. [32]. First, the acceleration signal is filtered using a discrete wavelet transform (DWT) with a Daubechies order 2 (db2) wavelet across six levels, focusing on levels 6 and 7. Gait frequency is estimated from prominent peaks in the filtered signal along the vertical (y) and forward (x) axis (Fig. 4Bi). A Morlet order 1 (morl) CWT is then applied to the velocity signal, obtained by integrating the acceleration signal, along the lateral (z) axis, using the estimated gait frequency as the input scale. Those wavelets were chosen for their similar shape to the signal as well as their ability to highlight the characteristics of interest related to gait [33]. Each footstep is identified as a local maxima or local minima in the resulting CWT signal (Fig. 4Bii), corresponding to right or left steps, respectively. The first gait segment begins with the first footstep following the maximum flexion point of the sit-to-stand phase and ends at the onset of the turning phase. All wavelet types used in this process were selected for their robustness in characterizing footstep waveforms. HS and TO events, defined as the initial foot contact with the ground and the final toe contact before swing, respectively, are detected using a similar CWT-based approach. Both events are extracted from the y velocity signal using wavelets with adaptive scales based on the estimated step frequency. Specifically, HS events are identified as local maxima in the db2-based CWT (Fig. 4Biii), while TO events correspond to local minima in the morl-based CWT (Fig. 4Biv). Also, smoothed velocity signals were used because they showed better robustness and precision than using the acceleration, which differs from the method proposed by Caramia et al. [32] and Del Din et al. [31]. Effectively, we study and compare carefully different wavelets on different signals to ensure the best results. Foot side classification for HS and TO is determined using the cycle CWT. The nearest footstep to a TO event defines its side. Conversely, the side of an HS event is assigned as the opposite of the nearest footstep. This approach ensures consistent and accurate step labeling across the gait cycle (Fig. 4Bii).Fig. 4Study overview. A Algorithm global overview. B Signal analysis for gait event. i Gait frequency computation from local minima detection. ii Footstep detection in morl CWT, obtained from z velocity signal. iii HS detection in db2 CWT, obtained from y velocity signal. iv TO detection from morl CWT, obtained from y velocity signal. In subfigures (ii), (iii), and (iv), circle markers represent left events and cross markers represent right events Study overview. A Algorithm global overview. B Signal analysis for gait event. i Gait frequency computation from local minima detection. ii Footstep detection in morl CWT, obtained from z velocity signal. iii HS detection in db2 CWT, obtained from y velocity signal. iv TO detection from morl CWT, obtained from y velocity signal. In subfigures (ii), (iii), and (iv), circle markers represent left events and cross markers represent right events Turn segmentation The algorithm developed for segmenting the turning phase is based on the angular velocity around the y axis, as measured by the gyroscope. The angular velocity signal is first filtered using a fourth-order low-pass Butterworth filter with a cutoff frequency of 10 Hz to reduce high-frequency noise, which targets the human movement only [34]. The algorithm then identifies the two most prominent peaks in the absolute value of the filtered signal, which correspond to the primary turning movements. To determine the precise start and end points of the turning phase, the method incorporates the iTUG model proposed by Salarian et al. [35]. This model estimates turning moments and related parameters by fitting a mathematical model to the integrated gyroscope signal around the y axis, thereby enhancing the accuracy of turn phase detection.Fig. 5Turn segmentation Turn segmentation TUG-VIMU spatio-temporal parameters assessment Spatio-temporal parameters refer to specific metrics that reflect the quality of movement, offering valuable insights into potential mobility deficits and aiding in clinical interpretation. Because these parameters are closely tied to the nature of the movement being performed, they are inherently specific to each phase of the TUG test. STS is primarily characterized by two STPs commonly reported in the literature: the transfer duration and the trunk lean angle, the latter defined as the maximum x flexion reached during the STS phase. To compute the trunk lean angle, the angular velocity around the z axis—measured by the gyroscope—is first integrated to obtain the trunk angle signal. The maximum lean angle is then calculated as the difference between the initial angle while seated and the peak flexion angle reached during the STS movement, as described by Salarian et al. [36]. Gait is first characterized by velocity, which reflects the speed of displacement. This metric is widely recognized in the literature as a global indicator of mobility [37]. In this study, the gait velocity of each segment is assessed by computing the difference between the first and last ArUco positions, divided by the corresponding time. The test gait speed is then defined as the mean velocity of both segments. Another key parameter is gait cadence, defined as the number of steps taken per minute. It is computed as the ratio of the number of steps performed during the gait phase by the duration of that phase, scaled appropriately to yield steps per minute. In addition to these global parameters, gait detailed STPs include step length, double support time, and step duration. Temporal indicators can be directly computed using gait events: step duration corresponds to the time elapsed from the contact of one foot to the contact of the opposite foot, while double support time corresponds to the ratio between the time elapsed with both feet on the ground to the step duration, in percentage. Step length corresponds to the distance travelled between two consecutive HS, in the direction of displacement or x axis. It is specifically defined as the norm of the vector connecting two consecutive HS points, based on the ArUco marker’s translation data. Step length is the distance travelled between two consecutive HS events, measured along the direction of displacement (i.e., the x-axis). It is computed as the norm of the vector connecting two consecutive HS positions, based on the translation data from the ArUco marker. Turn STPs include duration, maximum speed, mean speed, and the number of steps taken during the turn. The maximum speed corresponds to the peak value in the velocity signal along the y-axis (see Fig. 5). Similarly, the mean speed represents the average velocity during the first turn. The number of steps refers to the total step count within the first turn, calculated using the TUG-VIMU algorithm based on the CWT, as described in the gait segmentation section (Fig. 4B). Validation method Participants The number of participants is calculated considering a one-sided hypothesis with a risk of type 1 error of 20% and a risk of type II error of 5%, as well as a standard deviation of the differences of 3 cm, estimated from preliminary development tests, we find:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{gathered} n = \frac{{\left( {Z_{\alpha } + Z_{\beta } } \right)^{2} }}{{S_{{2GD}}^{2} }} \hfill \\ = \frac{{\left( {Z_{\alpha } + Z_{\beta } } \right)^{2} \sigma _{{diff}}^{2} }}{{\mu _{{diff}}^{2} }}\left. {\begin{array}{*{20}l} {Z_{\alpha } = 1,645} \\ {Z_{\beta } = 0,842} \\ \end{array} } \right\} \hfill \\ n = \frac{{\left( {1,645 + 0,842} \right)^{2} 3^{2} }}{{2^{2} }} = \frac{{0,185 \cdot 3^{2} }}{{2^{2} }} = 13,9 \hfill \\ \end{gathered}$$\end{document}In addition, a 10% margin is added to anticipate problems that could prevent continuation with a participant. This 10% margin is sufficient, since the project presents few problems expected, due to the participant’s level of involvement, which is considered minor (a single session of approximately 2 h). Thus, a total of 16 healthy adult participants were enrolled in the study (56% ; mean age: 45 ± 19 years; 8 over 60 years; mean body mass index: 23.1 ± 3.8; mean height: 1.74 ± 0.09 m). Inclusion criteria required participants to be 18 years of age or older. Exclusion criteria included any mobility impairments that could interfere with completing the protocol or significantly affect gait patterns. The study was conducted at the Research Centre on Aging in Canada and was approved by the institutional Ethics Committee. All participants provided written informed consent before participation. Protocol Participants were equipped with a GoPro camera mounted at the trunk using the GoPro Performance Chest Mount strap. Additionally, 39 passive markers were placed on anatomical landmarks following the specifications of OptiTrack’s conventional full-body model (version 3.1) [38]. This setup enabled the use of an optoelectronic gold-standard motion capture system. Marker trajectories were recorded using a network of 17 PrimeX cameras, consisting of 8 PrimeX 13 cameras (1.3 MP resolution) and 9 PrimeX 41 cameras (4.1 MP resolution). Prior to data collection, the camera system was calibrated to achieve sub-millimeter accuracy. Marker positions were captured at a sampling rate of 100 Hz and reconstructed into 3D coordinates using Motive software (version 3.0.3) [39]. Participants were then instructed to perform repeated 3-m Timed-Up-And-Go (TUG) tests at three self-selected gait speeds: slow, regular, and fast. To test the robustness of the algorithm to variations, participants were asked to walk slower or faster than their regular speed. To minimize potential fatigue effects, the order of speed conditions was counterbalanced across participants using a Latin Square design. Each participant completed a total of 18 TUG trials. A brief intermission was provided at the midpoint of the session, during which the GoPro camera was temporarily removed. Each trial began with the participant seated in a chair. Upon a start signal, the participant stood up, walked 3 ms, turned around a cone, and returned to the initial seated position (Fig. 1A) [2, 12]. Mocap data processing This section outlines the methods used to segment and extract STPs from the Mocap data across the sit-to-stand, gait, and turning phases. Mocap segmentation The MoCap-based segmentation algorithm identifies the STS phase using the first and last local maxim in the vertical displacement of the sternum marker, as described by Beyea et al. [40]. Gait segmentation was performed using the RDSF algorithm, which identifies gait phases based on HS and TO events [10]. The algorithm was applied to the realigned signal (see Sect. 2.2) to detect these key events. Each gait segment was defined as beginning at the first HS and ending at the last TO prior to the turning phase. Turn segmentation was performed using the yaw angular velocity derived from trunk 3D marker trajectories. The filtering and peak detection procedures mirrored those used in the TUG-VIMU segmentation approach. GS STPs Spatio-temporal parameters from the Mocap system were computed for the sit-to-stand, gait, and turn phases. STS: The lean angle was calculated as the maximum angle between the sternum and hip segment relative to the upright standing posture, following the method described by Salarian et al. [36]. Gait: HS and TO events were identified during the segmentation process. Step length was then computed using the x trajectory of the foot markers, as described in Sect. 2.3.2. Turn: Turn-related STPs included duration, maximum speed, mean speed, and number of steps. The first three parameters were computed using the same principles applied in the TUG-VIMU method. Step count during the turning phase was determined using the RDSF algorithm, as detailed in the previous subsection. Validation analysis To address the study objectives, statistical analyses were conducted to evaluate the precision and reliability of the TUG-VIMU system compared to the gold-standard motion capture system. Specifically, we assessed the accuracy of segmentation and derived spatio-temporal parameters, as well as the system’s robustness to repositioning. All analyses were performed using MATLAB R2023a (MathWorks, USA) libraries. Precision analysis The precision of TUG-VIMU was assessed by comparing every clinical metric to those obtained from the Mocap system. Acceptable error margins for each parameter were defined based on literature and clinical relevance (minimal clinical importance difference (MCID) or minimal detectable change (MDC)), as summarized in Table 1.Table 1Parameter’s objective errorPhaseParameterUnitObjectiveSourceTestDurations±3.40[41]STSDurations±0.47[42]Lean Angledegree±10.00[43]GaitHSs±0.05[44]TOs±0.05[44]Trunk Displacementm±0.05[45]Velocitym/s±0.08[46]Cadencestep/min±2.90[47]Step Durations±0.10[48]Step Lengthm±0.05[45]Double Support%±4.2[49]TurnDurations±0.40[42]Mean Velocitydeg/s±10.40[50]Maximum Velocitydeg/s±10.40[50]Number Step%±25.00[51] Table 1Parameter’s objective errorPhaseParameterUnitObjectiveSourceTestDurations±3.40[41]STSDurations±0.47[42]Lean Angledegree±10.00[43]GaitHSs±0.05[44]TOs±0.05[44]Trunk Displacementm±0.05[45]Velocitym/s±0.08[46]Cadencestep/min±2.90[47]Step Durations±0.10[48]Step Lengthm±0.05[45]Double Support%±4.2[49]TurnDurations±0.40[42]Mean Velocitydeg/s±10.40[50]Maximum Velocitydeg/s±10.40[50]Number Step%±25.00[51] Parameter’s objective error A non-inferiority t-test (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\alpha = 0.05$$\end{document}) was used to evaluate whether the error between systems remained within predefined acceptable limits for segmentation and gait events (HS and trunk displacement). To isolate spatial accuracy, temporal alignment was performed by synchronizing the TUG-VIMU data to the Mocap system using HS events, when applicable. Trunk displacement from TUG-VIMU was compared to that derived from the Mocap sternum marker trajectory between the HS event, so, we could isolate the measurement from the ArUco detection feature. Those parameters were chosen because of their high occurrence and because every other parameter arose from them. Robustness and reliability analysis To assess robustness, we compared TUG-VIMU outputs before and after a brief pause during which the GoPro camera was repositioned. This test-retest scenario allowed us to evaluate the system’s reliability from a clinical usability perspective. Reliability of parameters was quantified using Pearson’s correlation coefficient (r) on step length and step duration during gait, calculated from the mean error across trials for each condition. These parameters were chosen because they include multiple events, allowing the computation of a mean error. This test assesses how systems consistently vary with each other, corresponding to a relative reliability. The Table 2 presents thresholds that were used to interpret the strength of the correlation, based on guidelines from Koo and Li [52]. A minimum correlation of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$r = 0.75$$\end{document} is deemed satisfactory for this study.Table 2Pearson’s correlation coefficient (r) thresholdrReliability0.00 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\le$$\end{document} |r| < 0.25Low0.25 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\le$$\end{document} |r| < 0.50Moderate0.50 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\le$$\end{document} |r| < 0.75Good0.75 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\le$$\end{document} |r| \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\le$$\end{document} 1.00Excellent Table 2Pearson’s correlation coefficient (r) thresholdrReliability0.00 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\le$$\end{document} |r| < 0.25Low0.25 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\le$$\end{document} |r| < 0.50Moderate0.50 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\le$$\end{document} |r| < 0.75Good0.75 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\le$$\end{document} |r| \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\le$$\end{document} 1.00Excellent Pearson’s correlation coefficient (r) threshold The detection rate of steps is used to assess the reliability of a test. It represents the number of true events detected divided by the total of events detected. Agreement analysis To further assess the agreement between systems, Bland–Altman (BA) plots were generated for each STP (Fig. 6). These plots were based on the mean values across trials and were used to visualize the level of interchangeability between TUG-VIMU and Mocap measurements. Results A grand total of 288 trials were collected. Of those, 35 trials were excluded due to poor test execution (e.g., excessive chair movement, marker loss) or algorithm errors caused by bad segmentation, insufficient steps before turning, or too much movement on the chair. Consequently, 253 trials were included in the analysis: 89 at slow speed, 86 at regular speed, and 78 at fast speed. Segmentation precision Test duration was assessed with a precision of 0.36 s, while the detection of individual phases showed a mean segmentation error of less than 0.4 s compared to the Mocap reference (STS: 0.37; Turn: 0.15). Detailed results are presented in Table 3. A total of 1231 HS events (592 slow, 383 regular, and 256 fast) and 1083 TO events (514 slow, 336 regular, and 233 fast) were detected and analyzed. The accuracy of gait events detection, including means, standard deviations (SD), and confidence interval (CI) results, are detailed in Table 3 for each gait speed condition.Table 3Segmentation AccuracyVelocityMean error(Sd)95% CIGlobal SegmentationTUG Duration (s)Slow0.32(0.47)[0.22, 0.42]Regular0.43(0.47)[0.33, 0.53]Fast0.34(0.42)[0.24, 0.44]Combined0.36(0.46)[0.31, 0.42]Phases SegmentationSTS (s)Slow0.35(0.41)[0.29, 0.41]Regular0.39(0.33)[0.34, 0.44]Fast0.36(0.28)[0.32, 0.41]Combined0.37(0.34)[0.34, 0.40]Turn (s)Slow0.15(0.21)[0.12, 0.18]Regular0.13(0.14)[0.11, 0.15]Fast0.16(0.18)[0.13, 0.19]Combined0.15(0.18)[0.13, 0.16]Step Events DetectionHS (ms)Slow19.98(16.29)[18.65, 21.31]Regular17.94(17.23)[16.20, 19.68]Fast19.98(25.07)[16.93, 23.03]Combined19.35(18.82)[18.29, 20.40]TO (ms)Slow41.79(37.61)[38.54, 45.03]Regular35.69(25.00)[33.02, 38.37]Fast47.87(29.42)[44.10, 51.64]*Combined41.22(32.73)[39.28, 43.16]Trunk Displacement (cm)Slow2.44(6.83)[1.72, 3.17]Regular3.63(8.94)[2.22, 5.04]*Fast2.25(1.42)[1.90, 2.61]Combined2.75(7.15)[2.16, 3.34]* Does not fulfill requirements Table 3Segmentation AccuracyVelocityMean error(Sd)95% CIGlobal SegmentationTUG Duration (s)Slow0.32(0.47)[0.22, 0.42]Regular0.43(0.47)[0.33, 0.53]Fast0.34(0.42)[0.24, 0.44]Combined0.36(0.46)[0.31, 0.42]Phases SegmentationSTS (s)Slow0.35(0.41)[0.29, 0.41]Regular0.39(0.33)[0.34, 0.44]Fast0.36(0.28)[0.32, 0.41]Combined0.37(0.34)[0.34, 0.40]Turn (s)Slow0.15(0.21)[0.12, 0.18]Regular0.13(0.14)[0.11, 0.15]Fast0.16(0.18)[0.13, 0.19]Combined0.15(0.18)[0.13, 0.16]Step Events DetectionHS (ms)Slow19.98(16.29)[18.65, 21.31]Regular17.94(17.23)[16.20, 19.68]Fast19.98(25.07)[16.93, 23.03]Combined19.35(18.82)[18.29, 20.40]TO (ms)Slow41.79(37.61)[38.54, 45.03]Regular35.69(25.00)[33.02, 38.37]Fast47.87(29.42)[44.10, 51.64]*Combined41.22(32.73)[39.28, 43.16]Trunk Displacement (cm)Slow2.44(6.83)[1.72, 3.17]Regular3.63(8.94)[2.22, 5.04]*Fast2.25(1.42)[1.90, 2.61]Combined2.75(7.15)[2.16, 3.34]* Does not fulfill requirements Segmentation Accuracy * Does not fulfill requirements STP precision Precision of STPs for STS (lean angle), gait (cadence, velocity, step duration, step length, double-support time) and turn (mean and max velocity, number of steps taken during the turn) are detailed in Table 4, along with their statistical significance.Table 4TUG-VIMU STPs AccuracyVelocityMean error(Sd)Limits of agreementSit-to-Stand Transfer STPsLean Angle (deg)Slow\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}4.89(6.69)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}18.00, 8.22]*Regular\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}4.33(5.83)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}15.75, 7.10]*Combined\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}4.61(6.18)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}16.72, 7.50]*Gait STPsVelocity (m/s)Slow\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}0.00(0.02)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}0.03, 0.03]Regular0.00(0.02)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}0.04, 0.05]Combined0.00(0.02)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}0.04, 0.04]Cadence (steps/min)Slow\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}0.03(1.49)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}2.96, 2.90]*Regular0.65(3.54)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}6.28, 7.59]*Combined0.31(2.69)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}4.97, 5.59]*Step Duration (ms)Slow0.55(8.32)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}15.76, 16.86]Regular3.60(10.45)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}16.88, 24.08]Combined2.08(9.42)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}16.39, 20.53]Step Length (cm)Slow\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}0.75(1.02)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}2.75, 1.25]Regular\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}0.64(2.63)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}5.80, 4.52]*Combined\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}0.69(1.97)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}4.54, 3.16]Double-Support (%)Slow\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}1.82(3.75)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}9.17, 5.52]*Regular2.48(2.01)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}1.46, 6.41]*Combined0.25(3.70)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}7.00, 7.50]*Turn STPsMean Velocity (m/s)Slow0.03(2.77)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}5.39, 5.46]Regular0.08(2.72)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}5.26, 5.42]Combined0.06(2.70)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}5.24, 5.35]Max Velocity (m/s)Slow\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}1.79(4.27)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}10.15, 6.57]Regular\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}2.57(4.68)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}11.74, 6.60]Combined\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}2.18(4.42)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}10.85, 6.49]Number of Steps (%)Slow\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}0.23(0.93)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}2.05, 1.58]Regular0.74(6.24)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}11.48, 12.97]Combined0.26(4.41)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}8.39, 8.91]* Does not fulfill requirements Table 4TUG-VIMU STPs AccuracyVelocityMean error(Sd)Limits of agreementSit-to-Stand Transfer STPsLean Angle (deg)Slow\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}4.89(6.69)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}18.00, 8.22]*Regular\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}4.33(5.83)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}15.75, 7.10]*Combined\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}4.61(6.18)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}16.72, 7.50]*Gait STPsVelocity (m/s)Slow\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}0.00(0.02)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}0.03, 0.03]Regular0.00(0.02)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}0.04, 0.05]Combined0.00(0.02)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}0.04, 0.04]Cadence (steps/min)Slow\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}0.03(1.49)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}2.96, 2.90]*Regular0.65(3.54)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}6.28, 7.59]*Combined0.31(2.69)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}4.97, 5.59]*Step Duration (ms)Slow0.55(8.32)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}15.76, 16.86]Regular3.60(10.45)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}16.88, 24.08]Combined2.08(9.42)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}16.39, 20.53]Step Length (cm)Slow\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}0.75(1.02)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}2.75, 1.25]Regular\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}0.64(2.63)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}5.80, 4.52]*Combined\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}0.69(1.97)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}4.54, 3.16]Double-Support (%)Slow\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}1.82(3.75)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}9.17, 5.52]*Regular2.48(2.01)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}1.46, 6.41]*Combined0.25(3.70)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}7.00, 7.50]*Turn STPsMean Velocity (m/s)Slow0.03(2.77)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}5.39, 5.46]Regular0.08(2.72)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}5.26, 5.42]Combined0.06(2.70)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}5.24, 5.35]Max Velocity (m/s)Slow\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}1.79(4.27)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}10.15, 6.57]Regular\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}2.57(4.68)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}11.74, 6.60]Combined\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}2.18(4.42)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}10.85, 6.49]Number of Steps (%)Slow\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}0.23(0.93)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}2.05, 1.58]Regular0.74(6.24)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}11.48, 12.97]Combined0.26(4.41)[\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}8.39, 8.91]* Does not fulfill requirements TUG-VIMU STPs Accuracy * Does not fulfill requirements System reliability Reliability of TUG-VIMU is characterized by a 98% step detection during gait phases. Pearson’s correlation coefficient, assessing relative reliability between TUG-VIMU and GS, is above 0.87 for step length, and greater than 0.78 for step duration at slow and regular gait speed, as detailed in Table 5.Table 5Pearson’s correlation coefficient (r) of clinical parametersVelocityrReliabilityStep LengthSlow0.93ExcellentRegular0.89ExcellentFast0.87ExcellentStep DurationSlow0.98ExcellentRegular0.78ExcellentFast0.47Moderate Table 5Pearson’s correlation coefficient (r) of clinical parametersVelocityrReliabilityStep LengthSlow0.93ExcellentRegular0.89ExcellentFast0.87ExcellentStep DurationSlow0.98ExcellentRegular0.78ExcellentFast0.47Moderate Pearson’s correlation coefficient (r) of clinical parameters Systems agreement for clinical assessment To assess the agreement between the TUG-VIMU system and Mocap for clinical evaluation, we compared key spatio-temporal parameters relevant to functional mobility. These included: step duration, step length, double support (DS), cadence, gait velocity, turn duration, mean and maximum turn velocity, sit-to-stand (STS) duration, and lean angle. A total of 568 steps (158 regular, 346 slow, and 64 fast), 413 DS phases (104 regular, 254 slow, and 55 fast), and 253 data points for each of the remaining parameters were analyzed. Bland–Altman (BA) analyses for step duration, step length, and DS are presented in Fig. 6, illustrating the level of agreement between systems. Detailed biases, confidence intervals, and error metrics for all parameters are provided in Tables 4.Fig. 6Bland-Altman plots of clinical parameters. A Bland-Altman plot of gait velocity. B Bland-Altman plot of step duration. C Bland-Altman plot of step length. D Bland-Altman plot of percentage of DS Bland-Altman plots of clinical parameters. A Bland-Altman plot of gait velocity. B Bland-Altman plot of step duration. C Bland-Altman plot of step length. D Bland-Altman plot of percentage of DS Discussion This study aims to validate the performance of the TUG-VIMU system, an instrumented TUG test using a combination of a GoPro camera and ArUco markers. Its ability to detect key TUG phase events and compute spatiotemporal parameters was compared against a Mocap in a healthy population. Global system performance Overall, the TUG-VIMU demonstrated high temporal precision in detecting the test duration, as well as the time of each phase. The average error (Table 3) of segmentation falls within the typical range of human reaction time, suggesting that the system achieves a level of precision comparable to or exceeding that of manual timing with a standard chronometer. These results are encouraging, as inaccurate segmentation can introduce bias into all other parameters. Regarding gait event detection, TUG-VIMU demonstrated high validity, achieving a 98% detection rate for both toe-off (TO) and heel-strike (HS). This indicates that the wavelet-based approach is robust across varying gait speeds, even when changing direction while walking (i.e., turn). Previous work by Del Din et al. [31] achieved similar detection rates using wavelets; however, their method was limited to a specific gait speed or required manual parameter tuning. In contrast, the TUG-VIMU method automatically estimates gait speed and dynamically adjusts wavelet parameters, enhancing its robustness and adaptability across different walking conditions. The system also demonstrated high temporal accuracy in gait events detection, with mean errors within the predefined 50 ms objective. As a point of comparison, the mean HS detection error of 19.35 ms falls within the variability observed when switching between different algorithms used to process optical motion capture data [10]. Looking into variations in precision between gait velocity, it is noted that TO detection was less reliable at fast and very slow speeds. At a very slow speed, the gait motion is altered by small steps, practically lifting the whole foot rather than pushing off the toes before. This kind of motion is important to consider because it can be frequently seen in a pathological population. These techniques to detect TO events from wavelets with a change of motion thus warrant further investigation. HS events were generally detected more accurately than TO, and both event types showed the lowest error at preferred walking speed. This observation is in concordance with Caramia et al. [53]. TUG-VIMU STP precision and reliability Instrumentation of a clinical performance test, such as the TUG, aims at enabling specific parameters related to the quality of movement or potentially linked to a mobility deficit to be captured objectively. With that in mind, this study investigated the ability of the TUG-VIMU system to assess the most commonly used spatio-temporal parameters traditionally measured during an instrumented TUG, focusing on their precision, their relative reliability compared to the same parameters assessed with a GS, and the impact of gait speed on their precision. However, the fast velocities are less representative of the population for which this test is being used. As such, the focus of this discussion will be given to slow and normal gait speed parameters to test the robustness of the algorithm to variations. In this context, the lean angle during the sit-to-stand transfer was identified with a mean accuracy within 5 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$^{\circ }$$\end{document}. Typically, the accuracy one can expect from a goniometer is generally accepted to be 10\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$^{\circ }$$\end{document}. Thus, the TUG-VIMU system is well beyond this point but could have higher accuracy compared to the study from Roetenberg et al. [54] who report 3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$^{\circ }$$\end{document} using wearable sensors on the back. Gait STPs included cadence, gait speed, step length, step duration, and double-support time. All parameters present mean errors below the Table 1 objective. For example, the mean error of step length assessment is below 0.69 cm despite the variation assessed within the parameter. Similarly, gait speed (error < 0.01 m/s), cadence (error < 0.5 steps/min), step duration (error < 1.5 ms), and double-support (error < 0.25%) are all deemed below the desired threshold. Relative reliability of step length and duration, assessed using Pearson’s correlation coefficient, all reached an excellent level of reliability (r \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$>=$$\end{document} 0.75) for slow and regular gait speed. Step length results show higher reliability than other methods using inertial sensors placed on the foot, such as the approach presented by Amitrano et al. [55]. This suggests that TUG-VIMU achieves better results than foot-mounted inertial systems, which are typically considered the most accurate for measuring step length. Likewise, step length estimates achieved the 5 cm objective on average. However, TO detection at fast speeds and step length estimation at regular speeds failed to reach statistical significance, likely due to a limited sample size resulting in a lack of statistical power. Clinical agreement To go one step further with the system clinical agreement, BA analyses were conducted for key parameters. This type of analysis allows to predict the 95% confidence interval one can expect if switching between the two compared systems. In this specific study, this will allow us to assess the variation between the gold standard and the TUG-VIMU. Step duration showed strong agreement with the GS, with a global bias of 2.08 ms and a confidence interval ranging from -\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$-$$\end{document}16.39 ms to 20.53 ms. As shown in Fig. 6. A, variability increased for shorter durations, with a slight negative bias at longer durations. Importantly, the highest accuracy was observed at slow walking speeds, which is clinically relevant for populations with mobility impairments. Step length also demonstrated high clinical agreement, except at regular speed in the younger adult population, where the confidence interval exceeded the predefined threshold. However, the older population met the objective across all speeds. Figure 6 shows that larger steps were associated with increased absolute error, possibly due to pendulum-like motion from taking longer steps to walk faster, and to the lower number of steps required to complete the test. DS was sensitive to TO detection errors, particularly at slow speeds where early TO detection led to underestimation. This trend is reflected in the BA plots. Nevertheless, another wavelet type could be explored to increase TO’s detection accuracy. Ji et al. [56] proposed a method to choose the right mother wavelet based on energy computation. However, their results are based on a wearable positioned on the legs. This technique could be a great way to improve TO accuracy. The STS lean angle assessment exhibited a mean error lower that 5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$^{\circ }$$\end{document}. However, the 95% confidence interval analysis has shown that accuracy can vary between −17\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$^{\circ }$$\end{document} to 8\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$^{\circ }$$\end{document}. This variation is hypothesized to be due to integration drift when the ArUco marker temporarily leaves the camera’s field of view. During these moments, lean angle estimation relied solely on gyroscope integration, which is prone to cumulative error. Maximum turn velocity also failed to meet the lower confidence interval threshold, likely due to signal filtering and the derivation of the roll-pitch-yaw (RPY) signal. Despite these exceptions, other temporal and spatial parameters, including test duration, turn duration, and gait speed, met both bias and confidence interval targets, supporting the clinical applicability of TUG-VIMU. These findings align with those of Amitrano et al. [55], who also reported reliable extraction of spatial information from inertial data, despite the inherent integration errors. Overall, TUG-VIMU demonstrates strong potential for clinical assessment, particularly in enabling precise quantification of key spatiotemporal parameters. For comparison, high-accuracy systems such as GaitRITE report step length errors of about 3.6 cm [48], a level of accuracy that TUG-VIMU reaches at slower walking speeds. In contrast, IMU-only approaches typically yield step length errors around 12 cm, which limits their usefulness. As a result, most IMU-based TUG assessments focus on temporal features of gait phase analysis and do not report spatial outcomes [57]. To further leverage the benefits of TUG-VIMU, the step length was computed using the inverted pendulum model proposed by Del Din et al. [31], and the estimates were compared against Mocap reference data (Appendix A). With a mean error around 9 cm, the accuracy did not meet the objectives of Sect. 3.4.1. These findings also demonstrate that TUG-VIMU provides a robust step length estimate, clearly outperforming the inverted pendulum model that underpins most IMU-only approaches and enabling reliable spatial parameters assessment. Limitations and future work This study focused on healthy participants, providing a solid foundation for evaluating the TUG-VIMU system under controlled conditions. Notably, the sample was intentionally balanced to include both younger and older adults, with half of the participants aged over 60, ensuring age-related variability in gait patterns. However, as all participants were healthy, future research should expand to include individuals with mobility impairments to further assess the system’s robustness across a broader spectrum of functional abilities and clinical presentations. From a technical perspective, the system currently experiences occasional loss of ArUco marker tracking, with an average loss of 19.3% per test, primarily caused by STS transitions and turns. It should be noted that the ArUco markers were used within their optimal range. However, this range may present limitations for other tests requiring detection at greater distances (e.g., a 10 m walk test). For instance, Neto et al. [22] reported higher detection performance within a range of 0.5 m to 4.5 m for markers measuring 6 cm \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\times$$\end{document} 6 cm. In these instances, spatial parameters are derived solely from IMU data, which may be more susceptible to integration drift. Despite these challenges, the integration of inertial sensors, a wearable camera, and ArUco tracking represents a novel and promising approach, with few comparable systems currently available. Additionally, the use of a cone as a turn marker in this study may influence turning strategies compared to floor tape, which allows for pivot turns. Future work could explore the impact of different turn markers on gait dynamics. Overall, the TUG-VIMU system shows strong potential for detecting gait events and computing clinically relevant parameters. Future enhancements may include real-time feedback, expanded visual marker integration, and validation in clinical populations. These developments could further establish TUG-VIMU as a valuable tool for clinicians monitoring mobility and rehabilitation progress in diverse patient groups. Conclusion Human movement kinematics analysis offers significant potential for identifying individuals at risk of falling, yet the widespread clinical adoption of gold-standard systems remains hindered by their cost, spatial requirements, and operational complexity. This study demonstrates that the TUG-VIMU system, an accessible, portable solution combining a wearable inertial unit and a single GoPro camera with an ArUco marker, is a viable alternative for instrumented TUG assessments. Our results demonstrate that TUG-VIMU delivers high temporal and spatial accuracy, reliable detection of gait phase events, and precise estimation of clinically relevant parameters, particularly in step duration and step length among older adults. These findings underscore the system’s potential to support objective gait analysis in routine clinical practice, especially in settings with limited resources. With its low cost, ease of deployment, and strong performance, TUG-VIMU stands out as a promising tool for enhancing mobility assessment and fall risk screening. Future work will aim to extend validation to clinical populations, improve marker tracking during occluded phases, and explore real-time feedback capabilities to further support clinical decision-making and patient monitoring.