Respiration is a crucial vital sign to indicate acidotic state of human body. In this paper, we have developed a Holo-Hilbert spectral analysis (HHSA)-based approach to detect subject's respiration frequency from wrist photoplethysmogram (PPG) signals. The HHSA is a two-layer EMD architecture which first decomposes the original signal into intrinsic mode functions and then tries to find additive and multiplicative interactions among the participating components. The HHSA enables frequency-modulated (FM) and amplitude-modulated (AM) features of a signal can be comprehensively presented on a Holo-Hilbert spectrum (HHS). With the help of HHSA, subject's respiration frequency can be identified on HHS by finding the AM frequency with the most prominent amplitude around the FM frequency of heart rate. The efficacy of the proposed method has been demonstrated in two designed experiments. In the first experiment, 75 subjects with ages ranged from 20-80 years old, and the difference between the detected results of proposed method and the readouts of transthoracic impedance plethysmography is 0.04 ± 0.96 breathes-per-minute (brm). In the second experiment, six subjects were requested to breathe at 6, 12, 18, 24 brm, following the pacing rhythms of a metronome. The difference between detected respiration rates and expected breathing rates is 0.01 ± 0.70 brm. The HHSA-based approach has manifested its capability to extract respiration-induced multiplicative component in PPG signal. It avoids the ambiguity of frequency representation in signal multiplication when using traditional additive decomposition methods.
ASJC Scopus subject areas
- Electrical and Electronic Engineering