ABSTRACT
Methods employed for the automatic detection of valve ailments from the heart records is a skilled task in cardiology. However, automated approaches, proposed for predicting the cardiovascular diseases greatly depend on the features extracted from the heart sound. The merits of time domain methods are its computational simplicity, less complexity and no domain transformation. In this paper, the reliability of employing a time domain feature namely, waveform length for the detection of murmur from heart signal is investigated on two well-known heart sound databases such as Pascal and physionet. The feature is statistically evaluated using Kolmogorov Smirnov test and the separability among the feature to distinguish murmur and normal is examined via histogram. The waveform length corresponds to normal and murmur showed a ‘P’ value of 5.37×10−07 and 5.59×10−08.
