[Paper Review] Wavelet Based Normal and Abnormal Heart Sound Identification using Spectrogram Analysis

Nilanjan Dey, Achintya Das+1 more|arXiv (Cornell University)|Sep 6, 2012

Phonocardiography and Auscultation Techniques6 references17 citations

TL;DR

This paper proposes a wavelet-based method for classifying normal and abnormal heart sounds using spectrogram analysis. By applying Discrete Wavelet Transform (DWT) to spectrograms and extracting row mean features, the approach achieves 82% classification accuracy using minimum Euclidean distance for classification, offering a promising tool for tele-diagnosis of cardiac conditions.

ABSTRACT

The present work proposes a computer-aided normal and abnormal heart sound identification based on Discrete Wavelet Transform (DWT), it being useful for tele-diagnosis of heart diseases. Due to the presence of Cumulative Frequency components in the spectrogram, DWT is applied on the spectro-gram up to n level to extract the features from the individual approximation components. One dimensional feature vector is obtained by evaluating the Row Mean of the approximation components of these spectrograms. For this present approach, the set of spectrograms has been considered as the database, rather than raw sound samples. Minimum Euclidean distance is computed between feature vector of the test sample and the feature vectors of the stored samples to identify the heart sound. By applying this algorithm, almost 82% of accuracy was achieved.

Motivation & Objective

• To develop a computer-aided system for identifying normal and abnormal heart sounds using spectrogram-based analysis.
• To leverage Discrete Wavelet Transform (DWT) for feature extraction from spectrograms rather than raw audio signals.
• To improve tele-diagnosis of heart diseases by enabling accurate, automated classification of heart sound patterns.
• To evaluate the performance of a spectrogram-based feature vector approach using minimum Euclidean distance classification.
• To demonstrate the feasibility of using DWT on spectrograms for reliable heart sound classification.

Proposed method

• The method uses spectrograms as the primary data source instead of raw audio signals.
• Discrete Wavelet Transform (DWT) is applied to the spectrograms up to n levels to decompose them into approximation and detail components.
• The approximation components from each DWT level are used to extract a one-dimensional feature vector by computing the row mean of each approximation subband.
• The feature vector of a test sample is compared with stored feature vectors using the minimum Euclidean distance criterion for classification.
• The classification decision is made based on the closest match in the feature space among the stored spectrogram samples.
• The approach relies on a database of spectrograms derived from heart sound recordings, not raw audio waveforms.

Experimental results

Research questions

• RQ1 Can spectrogram-based analysis using DWT improve the accuracy of normal and abnormal heart sound classification?
• RQ2 How effective is the use of row mean features extracted from DWT approximation components for heart sound classification?
• RQ3 Can a minimum Euclidean distance classifier achieve reliable classification when applied to spectrogram-derived features?
• RQ4 Does using spectrograms as input instead of raw audio signals enhance classification performance in heart sound analysis?
• RQ5 What level of classification accuracy can be achieved using this wavelet and spectrogram-based feature extraction method?

Key findings

• The proposed method achieved a classification accuracy of approximately 82% using the minimum Euclidean distance classifier.
• Feature extraction via row mean of DWT approximation components effectively captures discriminative patterns from spectrograms.
• Using spectrograms as the input data source, rather than raw audio, proved effective for classification tasks.
• The DWT-based decomposition of spectrograms enabled robust feature extraction across multiple frequency subbands.
• The method demonstrates potential for real-time tele-diagnosis of heart diseases due to its computational efficiency and accuracy.
• The results indicate that spectrogram-based analysis with wavelet transforms is a viable approach for automated heart sound classification.