Experimental results on FMCW radar based Human Recognition using only Doppler information

The demand of indoor human identification systems that prioritize privacy has led to an increased interest in wireless sensing devices such as radar. Several studies have identified distinctive characteristics of human movement that cause Doppler signatures, which can be used for subject identification. Following them, this paper proposes an alternative machine learning approach for human identification, which uses only Doppler information from each subject by using FMCW radar. Experimental results in the two-subject scenarios show that the proposed method can achieve an accuracy of 96.2% (±1.1%) when data for both training and verification are obtained from the same scenarios, while an accuracy of 67.0% (±2.8
^%
) can be achieved when data for the verification are obtained from different scenarios