Activity and intensity data for UWB radar classification

The task of automated activity classification has previously attracted various avenues of research, and has inspired different methodologies in solving the problem. We outline an unobtrusive method of detecting and classifying different activities and exercises using a 24 GHz UWB radar transceiver and a DNN. The radar transceiver module is used to record the data of a single individual carrying out 6 different activities within a closed environment, and the subsequently processed radar signals are used to train a CNN, which is used to classify the human activities and the intensity of the activities. 

Using a custom-designed experimental set-up, we measure 500 signal samples consisting of 6 different activities from each of the 7 participants using the UWB radar system. The dataset was recorded in a controlled environment and background noise was recorded prior to the experimentation and subsequently post-processed from the measurements. We define the methods used to record the activity data using the radar transceiver, and the techniques used to process the raw radar signals in this section using the denoising filter selection method.