Comparison of window shapes and lengths in short-time feature extraction for classification of heart sound signals

2026-04-16 · Source: cs.AI updates on arXiv.org · Field: Science & Research — Health & Medical Research, Artificial Intelligence & Machine Learning · Depth: Expert, quick

Summary

A study evaluated the impact of window shapes and lengths on short-time feature extraction for classifying heart sound (PCG) signals using a bidirectional long short-term memory (biLSTM) network. The research experimentally assessed three window shapes (Gaussian, triangular, rectangular) and three lengths for each, training and testing the biLSTM on extracted statistical features. Results indicated that the Gaussian window consistently yielded the best classification performance, with the triangular window performing comparably at a 75 ms length. The commonly used rectangular window proved to be the least effective choice. Specifically, a 75 ms Gaussian window achieved superior classification performance compared to a baseline method, highlighting the importance of window selection in PCG signal analysis.

Key takeaway

For Machine Learning Engineers developing diagnostic systems for cardiovascular pathology, selecting the appropriate windowing technique for PCG signal feature extraction is critical. Your models will achieve superior classification performance by utilizing a Gaussian window, particularly at a 75 ms length, over the commonly offered rectangular window. This choice directly impacts diagnostic accuracy and system reliability.

Key insights

Optimal window shape and length significantly enhance heart sound signal classification performance.

Principles

• Non-stationary signals benefit from short-segment feature extraction.
• Window spectral side lobes distort extracted features.
• Gaussian windows outperform rectangular for PCG signal splitting.

Method

An experimental evaluation compared three window shapes (Gaussian, triangular, rectangular) and three lengths for each, training a biLSTM network on statistical features extracted from PCG signals.

In practice

• Prioritize Gaussian windows for PCG signal segmentation.
• Consider 75 ms window length for optimal performance.
• Avoid rectangular windows for heart sound feature extraction.

Topics

• Heart Sound Classification
• Phonocardiography
• Bidirectional LSTM
• Short-Time Feature Extraction
• Windowing Techniques

Best for: AI Scientist, Machine Learning Engineer, Research Scientist

Related on AIssential

• Comparison of window shapes and lengths in short-time feature extraction for classification of heart sound signals — Gaussian windows and optimal lengths significantly improve heart sound classification with biLSTM networks.
• Shape and amplitude decoupling in pulsatile physiological signal synthesis and its evaluation — Pulsatile physiological signals, crucial for understanding cardiovascular dynamics, are often generated by methods that undesirably couple waveform shape and amplitude, hindering precise control.
• Lung-SRAD: Spectral-Aware Regularized Audio DASS with Dual-Axis Patch-Mix Contrastive Learning for Respiratory Sound Classification — Lung-SRAD uses State Space Models and spectral-aware regularization to improve respiratory sound classification by preserving high-frequency details.
• Phonetic Error Analysis of Raw Waveform Acoustic Models — Raw waveform models with BLSTMs and transfer learning improve phone recognition, with errors reflecting inherent phonetic similarities.
• Echo2ECG: Enhancing ECG Representations with Cardiac Morphology from Multi-View Echos — Aligning global ECGs with multi-view Echos resolves representational mismatch, yielding powerful, lightweight cardiac morphology features.
• Beyond Morphology: Quantifying the Diagnostic Power of Color Features in Cancer Classification — Pixel intensity alone, without morphology, provides a strong diagnostic signal for cancer classification.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.