Computer Assisted System for Predicting Human Behavior using Time Delay Neural Networks

The usage of wearable sensors in biomedical applications has led to generation of large stream of physiological sensor data that are time-stamped, continuous and rapid. These physiological data contains hidden knowledge that can be used for building a real-time decision making system for predicting and analyzing human behavior and health conditions. The objective of this work is to develop an effective classification framework that can help the physician in monitoring and analyzing human behavior. The proposed framework highlights the importance of using time delay neural network (TDNN) in constructing a classification model for physiological sensor data. The performance of the proposed framework is experimented and evaluated with the physiological sensor data acquired for human behavior analysis based on multimodal body sensing.