Audio-visual Recognition of Auscultatory Breathing Sounds using Fourier and Wavelet Analyses

The era of computer management of clinical data demands the establishment of new techniques to analyze auscultatory sounds that can be better understood by both physicians and patients. This paper describes visual-based recognition techniques of breath sounds using two spectrograms created using short term FFT and wavelet analyses. Changes of frequency, intensity, and tone with time of breath sounds (21 samples) were shown using spectrograms of two kinds. Consequently, abnormal breath sounds were simply detected by differences of those patterns at first sight. They assist the recognition of the associated condition of disease. We expect to become a diagnostic support system in the near future.