Temporal extension of heart sounds using synthetic spectrograms

The short duration of each cardiac cycle can make diagnosis of presence and type of heart murmurs difficult. One approach is to replay the heart sound at a slower rate. Although conceptually simple, this approach alters both the temporal and spectral content of the sound. By working in Time-Frequency (TF) rather than the time domain, such temporal/spectral limitations can be avoided. This paper presents such a method, temporally extending the heart sound in the TF domain and reconstructing the signal using an algorithm based upon the Minimum Least Squares (MLS) inversion of the Short-Time Fourier Transform (ST-FT), a commonly used TF representation. An example signal has been processed in this manner, namely, the atrial septal defect (ASD) murmurs. The temporally extended heart sounds allow for better appreciation of various acoustic features such as murmurs and splittings in the second heart sound.