Human Fall Detection Improvement Based on Artificial Neural Network and Optimized Zero Moment Point Algorithms

This paper presents the development of human fall protection system based on artificial neural network (ANN) and optimized zero moment point (ZMP) algorithms that can detect and protect falling people in real time. Evaluating the movement data of different parts of the body, the result shows that the double feet and waist make the most contributions. For the sake of monitor the motions of the feet and waist, the inertial MEME sensor-based hardware of the system was designed. The foot pressure measurement units and the Micro Inertial Measurement Units (μIMUs) was applied in this system with Zigbee network. In terms of improving the efficiency and accuracy of fall posture detection, the ZMP algorithm was optimized and combined with Artificial Neural Network. Experimental results showed that when combining the ZMP and artificial neural network algorithms together, the recognition of fall and ADL (Activities of Daily Life), the Sensitivity, the Specificity and the overall accuracy were all better than 98%.