Fuzzy SVM and IoT Technology for Improved Fall Detection in Individuals with Non-voluntary Movements

Falls pose significant health risks, both physically and psychologically. Accurate fall detection can reduce these risks and enhance the well-being of individuals with disabilities. Creating a reliable fall detection system is crucial for this purpose. Traditional fall detection techniques have relied on kinematic sensors, such as gyroscopes, accelerometers, vision cameras, and vibration sensors. These sensors capture kinematic data during a fall incident and employ Machine Learning (ML) algorithms, typically Support Vector Machines (SVMs), to determine if a fall occurred or not. However, some falls may exhibit data patterns resembling routine activities, especially in cases involving paralyzed individuals and their usual walking patterns. To address these challenges, this paper explores the fusion of Fuzzy Logic (FL) and SVM algorithms, known as Fuzzy SVM, with the integration of IoT technology to facilitate seamless connectivity and communication within the fall detection system. This approach aims to minimize false alarms and enhance accuracy by utilizing fuzzy membership functions to generate SVM outputs. This reduces labeling ambiguity, enhances SVM decision-making transparency, and equips SVMs to effectively handle uncertain information. As a result, fall detection becomes more precise and reliable, reducing the need for manual labeling and overcoming challenges associated with conventional SVM approaches. Notably, this research achieved 100% specificity, thereby enhancing system reliability and precision. The proposed method, incorporating a novel fuzzy membership in Fuzzy SVM, significantly reduced false alarms and improved fall detection performance compared to existing methods.