Abstract: Discrete wavelet analysis can efficiently be implemented by filter banks. This paper studies heart sound signals using a wavelet filter banks structure. The heart sound signals are viewed into different wavelet subspaces by expanding them uniquely into sub-band signals of varying time-frequency resolutions. This provides the ability to isolate and distinguish different vital time-frequency information from the original heart sound signal and to explore the physical characteristics of heart sounds and murmurs such as the presence of opening snap, third and fourth heart sounds, auscultatory gap , high and low frequency systolic and diastolic murmurs. Very specific important features can be extracted from this information and used as a base for an informative diagnostic tool to classify heart diseases. Preliminary results show the significance of this technique in emphasizing the diagnostic information obtainable from auscultation and in helping to bring cardiac auscultation into prominence again along with more modern technologies such as scans, echo and Doppler scans. These results are believed to be satisfactory and promising.