A Simple Signal Analysis Method Mitigating Estimation Errors of CARS

In this article, we propose a period estimation method suitable for the analysis of time-varying signals. It is known that frequency analysis methods such as fast Fourier transform (FFT) have a difficulty in analyzing low-frequency signals such as respirations and heartbeats, due to their low resolution in the low-frequency bands. To solve this problem, we have proposed the correlation of accumulation for real-time serial-to-parallel converter (CARS). CARS also enables us to analyze time-varying signals by applying the moving window on the time axis, as the short-time Fourier transform (STFT) technique. By applying the moving window, we can obtain a spectrogram as a result of signal analysis by CARS, i.e., a figure showing time versus frequency. However, a problem has been found that there are horizontal lines which do not rely on the analysis result appearing in the spectrogram obtained by CARS. This phenomenon prevents us to observe the CARS result. In this article, we propose a new method to eliminate the horizontal lines in the CARS results. Furthermore, we clarify that the proposed method improves the accuracy of CARS about 28.0%.