Human gait identification using UWB radar micro-Doppler signature

This study introduces the application of impulse radio ultra-wideband radar (IR-UWB) for person identification using the micro-Doppler signature of human gait. A signal processing framework is developed to extract the features from the spectrogram, including the physically interpretable features and the features derived through Principal Component Analysis (PCA). In addition, a cross-entropy-based approach was employed to quantify uncertainties in the radar classification model, enhancing reliability in the classifications. The efficacy of this method was evaluated using a dataset comprising radar data from 14 individuals, each recorded while walking for a duration of 10 minutes. Employing a Support Vector Machine (SVM) classifier, our approach achieved a remarkable accuracy rate of 97.1%. These results demonstrate the capability of the IR-UWB radar and our proposed algorithm to discern variations in gait patterns, thereby effectively identifying individuals. This indicates its potential as a valuable addition to smart home systems, significantly benefiting the development of home care systems.