Wireless Radar Breath Detection with Empirical Mode Decomposition Method

Physiological signal processing can be applied to emergency rescue and healthcare monitoring with an understanding of health status. Existing works have demonstrated the capacity of extracting respiratory signal with continuous-wave radar. Current breath detection adopts time-frequency transform and statistical pattern recognition method, which requires a lot of efforts to collect data. This paper proposes a method of respiratory detection that uses empirical mode decomposition for de-noising and I/Q (In-phase and Quadrature) Signal Demodulation. With domain-knowledge, the proposed method does not require a large dataset and processes signals in time domain to improve calculation efficiency. To verify the performance of the proposed method, experiments of detecting and recording breath signal from human participants were conducted. The accuracy of breath detection of the methods was obtained to assess performance. Waveshape distortion affects health monitoring judgement. To assess the degree of waveshape distortion of extracted respiratory signal, comparing waveshape between extracted signal and base signal was conducted. This finding could be used to aid the breath monitoring remotely at home to identify potential illness related to breath like apnea caused by brain.