(POSTER) Smart Shoe for Fall Detection Application
Md Shohanoor Rahman, Ali Nikoukar, Mesut Güneş · 2023
Real-time automatic fall detection is effective to mitigate the post-impact of fall injuries, especially in case of the elderly people as immediate detection transpires into a faster response. This paper proposes a methodology that utilizes subject's acceleration, angular velocity, pitch, and row as features and generates fall decisions based on processing and comparing feature data from multiple time frames. The developed algorithm uses features extracted from a single sensor unit consisting of 3 axis accelerometer and gyroscope integrated into shoes with processing units. In real-life evaluation, 5 healthy subjects performed 160 falls and 650 activities of daily living. Our methodology achieved a sensitivity of 94.38%, specificity of 98.58%, and accuracy of 96.30%. Evaluating using the SisFall public dataset sensitivity of 91.37%, specificity of 95.28%, and accuracy of 90.73% was achieved.