Feature extraction for heart sound recognition based on time-frequency analysis

Heart sounds provide clinicians with valuable diagnostic and prognostic information. Although heart sound analysis by auscultation is convenient as a clinical tool, heart sound signals are so complex and nonstationary that they are very difficult to analyze in the time or frequency domain. We have studied the extraction of features in the time-frequency (TF) domain for recognition of heart sounds through TF analysis, such as short time Fourier transform (STFT), Wigner distribution (WD), and wavelet transform (WT). After these transformations, we normalized the heart sound in the TF domain and compared a normal heart sound with an abnormal one to verify the clinical usefulness of our extraction methods for heart sound recognition