Mussell M J, Miyamoto Y
Department of Electrical and Information Engineering, Faculty of Engineering, Yamagata University, Yonezawa, Japan.
Front Med Biol Eng. 1992;4(2):73-85.
Respiratory sounds (RS)s were recorded from the trachea and chest of 10 normal adult subjects at respiratory air flow levels of 1.6, 2.1 and 2.6 l/s using an oral flow transducer, and at approximately 2.1 l/s without the flow transducer. Tracheal RS (TRS) and chest RS (CRS) frequency spectra were generated using Fast Fourier Transform, and the peak, mean and maximum frequency parameters were derived from each spectra. Parametric analysis showed: (i) all three parameters for TRS spectra are significantly higher than those for CRS spectra; (ii) TRSs are on average eight times louder than CRSs; (iii) both TRSs and CRSs are air-flow independent over the flow range, though TRSs are significantly modified by the flow transducer while CRSs are not; and (iv) though of similar loudness, inspired and expired RSs (both TRSs and CRSs) have some significant spectral differences. To compare the complex shapes of RS spectra, each spectra was divided into narrow frequency bands (to create a feature set) and principal component analysis was performed on all spectral feature sets. TRSs and CRSs were shown to be independent biological signals with little overlap in their respective spectral characteristics.