A Machine Learning Approach to Automatic Fall Detection of Soldiers
Leandro Soares, Gustavo Venturini, José Gomes, Jonathan Efigenio, Pablo Rangel, Pedro Henrique Gonzalez, Joel dos Santos, Diego Brandão, Eduardo Bezerra · 2025
Read A Machine Learning Approach to Automatic Fall Detection of Soldiers by Leandro Soares, Gustavo Venturini, José Gomes, Jonathan Efigenio, Pablo Rangel, Pedro Henrique Gonzalez, Joel dos Santos, Diego Brandão, Eduardo Bezerra (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-2501-15655.txt.
Introduction
Leandro Soares and co-authors detect falls among soldiers with wearable sensors in a 2025 paper. Troops in urban combat face high injury risk, and fast word of a severe fall can speed rescue. The work sits inside the Brazilian Navy's Soldier of the Future project. A casualty system aims to spot falls that may signal grave wounds and blood loss.
Smartwatches and phones collect motion data during drills and simulated falls. One-dimensional neural networks learn to tell true falls from normal moves. Sensor place on the body and choice of motion variables are both tested. Bayesian tuning sharpens the best model, which the authors discuss in detail.
Catalog record
| CREATOR | Leandro Soares, Gustavo Venturini, José Gomes, Jonathan Efigenio, Pablo Rangel, Pedro Henrique Gonzalez, Joel dos Santos, Diego Brandão, Eduardo Bezerra |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 71 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-15655
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A Machine Learning Approach to Automatic Fall Detection of Soldiers Leandro Soares, Gustavo Venturini, José Gomes, Jonathan Efigenio, Pablo Rangel, Pedro Henrique Gonzalez, Joel dos Santos, Diego Brandão, Eduardo Bezerra Abstract Military personnel and security agents often face significant physical risks during conflict and engagement situations, particularly in urban operations. Ensuring the rapid and accurate communication of incidents involving injuries is crucial for the timely execution of rescue operations. This article presents research conducted under the scope of the Brazilian Navy’s “Soldier of the Future” project, focusing on the development of a Casualty Detection System to identify injuries that could incapacitate a soldier and lead to severe blood loss. The study specifically addresses the detection of soldier falls, which may indicate critical injuries such as hypovolemic hemorrhagic shock. To generate the publicly available dataset, we used smartwatches and smartphones as wearable devices to collect inertial data from soldiers during various activities, including simulated falls. The data were used to train 1D Convolutional Neural Networks (CNN1D) with the objective of accurately classifying falls that could result from life-threatening injuries. We explored different sensor placements (on the wrists and near the center of mass) and various approaches to using inertial variables, including linear and angular accelerations. The neural network models were optimized using Bayesian techniques to enhance their performance. The best-performing model and its results, discussed in this article, contribute to the advancement of automated systems for monitoring soldier safety and improving response times in engagement scenarios. 1 Introduction Military personnel and security agents face physical risks in a variety of conflict and combat scenarios, both in urban and military contexts. In the military context, these scenarios include the retaking of military facilities, hostage rescue, target neutralization, and city defense, as well as United Nations peacekeeping missions. In urban operations, such as those carried out in low-income communities dominated by criminals in Brazil, the Guarantee of Law and Order (GLO) operations stand out. According to the GLO Manual [ 1 ] , GLO operations are conducted by the Armed Forces in serious situations of public order disruption when traditional public security forces are exhausted. In Brazil, these emergency operations are determined by the President of the Republic, delimiting a specific area and for a limited period. The goal of these operations is to ensure the preservation of public order, the safety of individuals, and the protection of property. Since 1992, the Brazilian Ministry of Defense has documented 146 GLO operations, 46 involving Brazilian Marines [ 2 ] . On the global stage, Brazil has played an active role in recent years, participating in over 30 United Nations peacekeeping missions and sending over 27,000 military personnel abroad for these missions [ 3 ] . A significant example is the MINUSTAH operation, which took place between 2004 and 2017, aimed at stabilizing Haiti after a civil conflict [ 4 ] . In this operation, the Brazilian Navy contributed naval forces and marine troops to support Haiti’s security and reconstruction. Additionally, it is important to mention the incursions of military police into low-income communities and slums in urban areas, which assume guerrilla-like characteristics and are part of daily life in Rio de Janeiro municipality. The Rio de Janeiro Military Police has recorded 477 police deaths in action since 2003 [ 5 ] . During those mentioned operations, soldiers are exposed to the risk of injuries from confrontations with enemy troops, with the most common being firearm injuries. Such injuries can lead to the soldier’s death if immediate communication of the incident with Command and Control (C2) is not established, and prompt rescue operations are not carried out. The first ten minutes after the injury are critical, essential for stabilizing the victim, and preventing the worsening of their condition. Procedures performed during this period, known as the “Platinum 10”, often save lives. Additionally, the first hour following the incident, known as the “Golden Hour”, is the crucial time in which the victim has the highest chances of survival, provided he/she receives medical care within this period [ 6 ] . Given the above considerations, it is vital for maintaining the life of a wounded soldier that the incident information reaches C2 quickly and accurately, allowing for the swift mobilization of the rescue operation. Hence, it becomes necessary to research and develop methods for the immediate detection of a soldier’s incapacitation and the swift transmission of this information to the command and control system. Considering the active participation of its soldiers in military operations that may result in armed conflicts, as well as the operations of other Armed Forces and Military Police, the Brazilian Navy, through IPqM, initiated the project called “Soldier of the Future”. Part of this project’s scope is the study of biometric parameters that can be collected from soldiers in action, aiming at the automatic inference of relevant incidents for support and/or safeguarding procedures. The research described in this paper is part of the “Soldier of the Future” project and aims to implement a Casualty Detection System. This work focuses on one component of this system, which is designed to identify injuries that could incapacitate a soldier and lead to significant blood loss. The system is composed of two models: (1) detection of the soldier’s fall based on inertial variables, and (2) classification of the fall as a result of hypovolemic hemorrhagic shock or an operational fall. This paper specifically focuses on the first model. Due to the nature of military operations, it is necessary for devices used for fall detection to be lightweight, easy to use, and easy to wear. Therefore, for the collection of biometric data and consequent fall detection, smartwatches and smartphones were chosen as portable wearable devices for collecting inertial variables. Data collection for the research involved 15 volunteer soldiers from the Brazilian Navy who performed various activities categorized as daily life activities, armed service activities, and military operations, in addition to simulated fall activities. Each participant wore a smartwatch on each wrist and a smartphone in the pocket of their uniform jacket, allowing for data collection from the right and left wrists as well as from an area near the body’s center of mass. This article presents the employed methodology, the obtained results, and provides a public dataset called IPqM-Fall at https://github.com/AILAB-CEFET-RJ/falldetection to support future research in this field. Artificial intelligence techniques, especially machine learning techniques such as artificial neural networks, have been widely applied to classification problems [ 7 ] . In this paper, we conducted fall detection experiments using 1D Convolutional Neural Networks (CNN1D). These networks were trained with data in both the time domain and the frequency domain. Additionally, the networks were trained separately with data from smartwatches worn on the right wrist, left wrist, and a smartphone positioned on the chest. The goal was to compare which location would be most suitable for sensor portability, considering predictive performance in fall classification. We also conducted experiments using only the magnitude of the linear acceleration of the accelerometer, as well as the x , y , and z components of this acceleration. The same procedure was applied to the angular acceleration of the gyroscope, and we experimented with an approach that combined the linear and angular acceleration