Human Activity Classification Based On Breathing Patterns Using IR-UWB Radar

In this paper, human activities were detected and classified in a non-contact method based on the respiration pattern of the human using impulse radio ultra-wideband (IRUWB) radar. To detect the respiration pattern of the human subject corresponding to different activities, a commercial IRUWB radar was employed. Firstly, a dataset of respiration signal using IR-UWB radar was created while the subject performed four different kinds of activities i.e. sitting idle (activity 1), kapalbhati pranayama (activity 2), speaking loudly (activity 3) and deep breathing meditation (activity 4). The respiration patterns of 20 subjects were collected while the subject performs different kinds of mentioned activities. After the respiration patterns are collected, the short-time Fourier transform (STFT) is applied to the received signal. Support vector machine (SVM) is used for the classification of human activities using the STFT of the respiration pattern as the feature. The method was able to achieve the best classification accuracy of 99 % for the activities 1 & 2, 92 % for the activities 1, 2 & 3 and 85.25 % for the activities 1, 2, 3 & 4.