Human Activity Recognition with a 6.5 GHz 6.5\text{\,}\mathrm{GHz} Reconfigurable Intelligent Surface for Wi-Fi 6E

Nuno Paulino, Mariana Oliveira, Francisco Ribeiro, Luís Outeiro, Pedro A. Lopes, Francisco Vilarinho, Sofia Inácio, Luís M. Pessoa · 2025

Read Human Activity Recognition with a 6.5 GHz 6.5\text{\,}\mathrm{GHz} Reconfigurable Intelligent Surface for Wi-Fi 6E by Nuno Paulino, Mariana Oliveira, Francisco Ribeiro, Luís Outeiro, Pedro A. Lopes, Francisco Vilarinho, Sofia Inácio, Luís M. Pessoa (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-14423.txt.

Introduction

Nuno Paulino and seven co-authors show how a smart wall panel can help Wi-Fi sense hand moves without cameras or worn sensors. Their study appeared in twenty twenty-five. The panel can steer radio waves in a set area, which sharpens what the waves pick up from a hand. The team built the panel to work at a high Wi-Fi band used by Wi-Fi 6E.

First the team checked that the panel steers waves as planned. Next they taped a data set for three hand signs and shared it in full. Two neural nets learned to sort the signs, one from mixed panel settings and one from tuned settings. Both scored above ninety percent right.

Catalog record

CREATORNuno Paulino, Mariana Oliveira, Francisco Ribeiro, Luís Outeiro, Pedro A. Lopes, Francisco Vilarinho, Sofia Inácio, Luís M. Pessoa
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE81 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-14423

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Human Activity Recognition with a 6.5 GHz 6.5\text{\,}\mathrm{GHz} Reconfigurable Intelligent Surface for Wi-Fi 6E Nuno Paulino, Mariana Oliveira, Francisco Ribeiro, Luís Outeiro, Pedro A. Lopes, Francisco Vilarinho, Sofia Inácio, Luís M. Pessoa Abstract HAR ( HAR ) is the identification and classification of static and dynamic human activities, which find applicability in domains like healthcare, entertainment, security, and cyber-physical systems. Traditional HAR approaches rely on wearable sensors, vision-based systems, or ambient sensing, each with inherent limitations such as privacy concerns or restricted sensing conditions. Recently, RF ( RF )-based HAR has emerged, relying on the interaction of RF signals with people to infer activities. RIS offers significant potential in this domain by enabling dynamic control over the wireless environment, thus enhancing the information extracted from RF signals. We present an HGR ( HGR ) approach that employs our own 6.5 GHz 6.5\text{\,}\mathrm{GHz} RIS design to manipulate the RF medium in an area of interest. We validate the capability of our RIS to control the medium by characterizing its steering response, and further we gather and publish a dataset for HGR classification for three different hand gestures. By employing two CNN models trained on data gathered under random and optimized RIS configuration sequences, we achieved classification accuracies exceeding 90 % 90\text{\,}\mathrm{\%} . I Introduction HAR is a growing field dedicated to identifying and classifying all forms of human activity, encompassing both static and dynamic motion [ 1 ] . Figure 1 illustrates different types of human activity, which include static activities, dynamic activities, and activities involving postural transitions [ 2 ] . Static activities include lying, sitting, and standing, which are generally easier to identify but can be confused with similar postures. Dynamic activities involve continuous movement, such as walking or running, which require distinguishing between different motion patterns. Activities with postural transitions, such as sitting to standing or walking to jogging, involve shifts between static and dynamic states and can be particularly complex due to overlapping motion patterns. The capability of performing accurate HAR has the potential to significantly impact various domains [ 3 ] . In healthcare, HAR can assist in monitoring patient conditions and detecting early signs of medical issues [ 4 ] . In smart environments, HAR can enhance home assistance services to improve quality of life and autonomy [ 5 ] . In sports, it can contribute to improving personal health and well-being [ 6 ] . In security, HAR can be used for surveillance to identify unusual events [ 7 ] . Additionally, HAR improves human-computer interaction (HCI) by enabling more intuitive and responsive interfaces [ 8 ] . Fig. 1: Types of Human Activity Recognition and respective examples. However, HAR is complex, requiring selecting and deploying suitable sensors to collect information-rich data, pre-processing the data, and extracting relevant features. Afterward, a range of approaches can be considered to process the data, including machine learning and deep learning algorithms to recognize activities [ 9 ] . Regarding specific sensor types, HAR approaches may be vision-based, through monocular or stereo cameras [ 10 ] , based on sensors such as accelerometers, which can be worn on the body or integrated into smartphones or smartwatches to capture direct movement data [ 11 ] . Additional sensors can provide secondary parameters like temperature and humidity, and provide further contextual information [ 12 ] . Another emergent method for HAR is the use of RF characteristics of the environment, where data relative to the interaction of stray or controllable RF signals with objects or people within a given SoI ( SoI ) is used as the sensor component of the system. By relying on RF signals, activities can be detected or classified through the effects of body movements on the propagation of these signals, allowing for HAR techniques that address concerns about monitoring and privacy [ 13 ] . It does not require wearable or environmental sensors, providing a noninvasive and lower cost approach, instead exploiting the already existing plethora of RF signals in the environment that originate from ubiquitous commodity networks like 5G or WiFi. Since reflected RF signals are also captured by receiving antennas, information can be gathered even in NLoS ( NLoS ) conditions [ 14 ] , and depending on the signal frequency, even through-wall detection can be considered [ 15 , 16 ] . Finally, the use of higher frequency signals can even provide a high spatial resolution, which has demonstrated advantages over vision-based systems for multi-person recognition [ 17 ] . Despite these advantages, wireless environments are complex and difficult to model. Not only is there no direct control over the stray RF signals in a region of interest, but there is also the complexity of multi-path interference, caused when signals reflect off different surfaces and reach the receiver through various paths [ 18 ] . Thus the difficulty of clearly relating and modeling the parameters of received RF signals to the status of the environment derives from a largely uncontrolled propagation medium. (a) Regular Refletor (b) RIS Fig. 2: Regular reflector vs. RIS , which achieves non-specular reflections through programmable phase shifts (adapted from [ 19 ] ). RIS ( RIS ) are a promising solution to achieve this medium control. They are composed by a 2D matrix of antenna elements, also called unit cells. These cells contain no active RF chains and act only as reflectors. Simple electronics are used to change the reflection coefficient of each individual unit cell, and consequently a RIS can adjust the surface impedance to produce specific phase shifts for each of the reflected replicas of the incident signal, thus achieving control of the wireless environment [ 20 , 21 ] . For instance, whereas a specular reflector like a metal surface shown in Figure 2a deflects a signal at an angle equal to the angle of incidence, specific control patterns applied to the unit cells of a RIS can be used to generate constructive interference to focus the total reflection into a particular direction, achieving non-specular reflections, as shown in Figure 2b . The advantages for communication systems are increased spectral efficiency and power consumption, by reducing radiation towards unintended directions, and support for NLoS communications, since links can be redirected around obstacles. But applications go beyond communication, as this medium control places RIS as a candidate technology to also enable applications like localization, imaging, or sensing [ 22 , 23 , 24 , 20 , 25 , 26 ] , in emergent networks like 6G [ 27 ] . Concurrently, the WiFi-6E standard has recently been defined, promising to deliver a throughput of up to 9.6\text{\,}\mathrm{G}\mathrm{b}\mathrm{p}\mathrm{s} by extending WiFi capability into the 5.925\text{\,}\mathrm{GHz} to 7.125\text{\,}\mathrm{GHz} range. This expansion allows for up to seven 160\text{\,}\mathrm{MHz} MHz channels, compared to the two channels available in Wi-Fi 5 and Wi-Fi 6, or alternatively up to 59 channels with a bandwidth of 20\text{\,}\mathrm{MHz} . Despite this, this frequency range suffers from greater path loss relative to previous WiFi standards, and also greater attenuation through walls or objects, compromising NLoS operations. This introduces the need for signal repeaters or greater transmission power. Considering the signal manipulation capabilities explained before, namely signal steering, RIS are once again placed as a

Read the complete text