Cosine Transform with Frequency-Domain Correction Technique for Radar-Based Short-Time Vital Sign Monitoring

This research focuses on improving the short-time accuracy of non-contact vital sign monitoring using quadrature self-injection-locked (QSIL) radar. Various methods of signal decomposition such as filtering, variational mode decomposition (VMD), and discrete wavelet transform (DWT) in short window lengths are discussed to compare the errors in time and frequency domain. Based on the comparison result, the filtering method is adopted in this work due to it having the lowest distortion level in short window lengths. Apart from using Fourier-transform-(FT)-based IQ demodulation methods, such as arctangent demodulation (AD) or complex signal demodulation (CSD), this paper proposes a cosine transform (CT) with a frequency-domain correction technique to improve the result of frequency estimation for short-time heart rate (HR) monitoring with a window length of 2-5 seconds. In a comparison with the FT-based methods, the experiment conducted on 5 subjects for a duration of 60 s for each subject shows that the proposed method had an average error of 5.6% in a 3-s window length.