Enhanced Frame-Similarity Technique for Detection of Respiratory and Cardiac Rhythms in Auscultation Records

An enhanced algorithm for recognizing the rhythms associated with cardiac and respiratory activity in auscultation signals has been developed on the basis of previously proposed approaches. The essence of this technique is to evaluate the energy similarity of fragments of the original record at their superposition with a progressive time shift. To estimate the time evolution of signal frequency content, the Short-Time Fourier Transform is applied to records obtained by the developed portable one-channel digital auscultation complex. The free terms of the linear regression model for “instant” amplitude-frequency spectrum calculated in four information-significant octaves with central frequencies of 63, 125, 250, and 500 Hz are chosen for a measure of acoustic energy variation. The procedure of averaging of initial sample at one-second time intervals is used to segregate the cardiac and respiration periodicity into “fast” and “slow” channels. It is shown that search for minimums of the introduced distances depicting the similarity of compared fragments provides a reliable tool for detecting of life activity rhythms contained in the auscultation signal.