Phonocardiogram signal compression using sound repetition and vector quantization.

Area of Science:

• Biomedical Engineering
• Signal Processing

Background:

• Phonocardiogram (PCG) signals are crucial for long-term heart monitoring.
• Large PCG data volumes necessitate compression for storage and telemedicine transmission.
• Heartbeat cyclicity offers opportunities for data compression by eliminating redundancies.

Purpose of the Study:

• To propose a novel PCG signal compression method.
• To exploit the cyclical nature of heartbeats for efficient data reduction.
• To reduce data storage and transmission requirements for PCG signals.

Main Methods:

• A two-stage compression process: training and compression.
• Training stage: Decomposing PCG signals into time-frequency components and creating a dictionary of basic components.
• Compression stage: Reconstructing sounds from the dictionary and compressing residuals using vector quantization with quick search parameters.

Main Results:

• Achieved compression ratios ranging from 8-149 depending on signal type (normal, median murmurs, heavy murmurs).
• Maintained a low distortion level of approximately 5% (root-mean-square difference).
• Demonstrated superior performance compared to existing compression methods via simulations.

Conclusions:

• Repetitive components in cyclical sounds can be effectively removed for compression.
• Vector quantization further reduces redundancies in the signal residual.
• The proposed method offers improved performance for PCG signal compression.