AUDIO-VISUAL BASED RECOGNITION OF AUSCULTATORY HEART SOUNDS WITH FOURIER AND WAVELET ANALYSES

Since there are various difficulties associated with auscultation techniques (e.g., the detection/recognition of murmurs and sound tone changes within approximately one second), we have proposed an audio-visual based technique to examine and visualize heart sounds for both physicians and patients. To overcome auscultation difficulties, the technique can be used to assist in the understanding of the heartbeat, the detection of heart disease, and the digital database management of the auscultation examination. Auscultatory sounds are visualized by both an FFT image and a wavelet image to detect any abnormal heart sounds. A simple technique of pattern classification has been established using short-time Fourier transform and wavelet analyses to detect abnormal sounds. This new technique is expected to be simple and practical in this era of computer-managed clinical data. The result indicates that there is a possibility of developing a fully automatic detection system based on a map of standard sounds in the near future. A simple auscultation method is also expected to be developed for in-home use.