FMCW Radar Sensor Based Human Activity Recognition using Deep Learning

Human Activity Recognition (HAR) has found many applications in several disciplines such as smart home and elderly healthcare units. The robustness of radar sensor against the environmental conditions make it a suitable candidate to recognize human activities. In this paper, we used Frequency Modulated Continuous Wave Radar (FMCW) radar for recog-nizing human activities in an unconstrained environment. Seven different activities are performed randomly at different distances from radar and a multi-class classification problem is formulated. Performed activates are recorded with single FMCW radar and a deep-learning classifier is used for recognition. The target range variations generated while performing the predefined human activates are fed as an input to the features extraction block of three Convolutional Neural Network and a softmax classification is performed. Overall recognition accuracy of 91% is achieved.