Karel Octavianus Bachri

Frequency Extraction of Phonocardiogram Signal using Fourier Transform

Introduction

Frequency extraction of phonocardiogram signal using fourier transform. Extract phonocardiogram (PCG) signal frequencies using Fourier Transform. Analyze normal and abnormal heart sounds, identifying dominant 50-150Hz & subdominant 450-650Hz.

Abstract

This article presents Fourier Transform application to extract features of Phonocardiogram Signals into its frequency components. Data was taken from Physionet Dataset of Phonocardiogram which comprises of normal and abnormal heart condition. Raw data was preprocessed using time clipping of 2 seconds at certain area that contains less noise. A lowpass filter was applied to denoise the raw signals. Experiments show the PCG of normal hearts has a dominant frequency of 50Hz to 150Hz, with the subdominant frequencies of 450 Hz to 650 Hz. The subdominant frequency of the normal hearts sometimes show anomaly with more amplitude compared to the dominant frequency.

Review

This article, "Frequency Extraction of Phonocardiogram Signal using Fourier Transform," tackles a highly relevant problem in biomedical signal processing: the application of Fourier Transform (FT) to extract meaningful frequency features from Phonocardiogram (PCG) signals. The overarching goal of distinguishing between normal and abnormal heart conditions using these features is crucial for advancements in non-invasive cardiac diagnostics. The authors' choice to utilize data from the well-established Physionet Dataset provides a robust foundation for their experimental analysis, and the initial preprocessing steps involving time clipping and lowpass filtering indicate a practical approach to handling real-world signal data. A significant contribution of this work lies in its systematic application of the Fourier Transform to identify specific frequency characteristics within normal PCG signals. The abstract details empirical findings, pinpointing a dominant frequency range of 50Hz to 150Hz and a distinct subdominant range of 450Hz to 650Hz for normal hearts. The observation that the subdominant frequency can, at times, exhibit a higher amplitude than the dominant frequency in healthy individuals is an intriguing "anomaly" that merits deeper exploration. These identified frequency ranges offer valuable preliminary features that could potentially serve as biomarkers for further classification or diagnostic algorithms. While the article presents a foundational application, several aspects could be elaborated upon to enhance its comprehensiveness and impact. A primary concern is the notable absence of specific findings for "abnormal heart conditions" in the abstract, despite their inclusion in the study's scope. Presenting how the frequency characteristics diverge in abnormal cases would be crucial for establishing the method's diagnostic utility. Furthermore, the selection criteria for "certain area that contains less noise" during time clipping could benefit from more objective justification. It would also be valuable to discuss the inherent limitations of a global Fourier Transform for analyzing non-stationary signals like PCG, perhaps by comparing its performance or suggesting advanced time-frequency analysis techniques (e.g., Wavelet Transform) that might better capture the transient nature of heart sounds. Finally, the "anomaly" of subdominant frequency amplitude variation could be explored with greater statistical detail and potential clinical correlations.

Full Text

Comments

You need to be logged in to post a comment.