Radar Data Cube Analysis for Fall Detection

Abstract

In recent years, radar has been employed as a fall detector, due to its superior sensing capabilities and penetration through walls. In this paper, we introduce a multi-linear subspace fall detection scheme that exploits the three radar signal variables: slow-time, fast-time, and Doppler frequency. The proposed approach attempts to find the optimum orthonormal subspaces that minimize the reconstruction error for different modes of the radar data cube. Experimental results based on real radar data obtained from multiple subjects and aspect angles demonstrate that the proposed multi-dimensional principal component analysis (MPCA) yields the highest overall classification accuracy among other methods including physically interpretable pre-defined features and spectrogram-based standard PCA.