.png){kind=icon}
.png){kind=icon}
Return
A novel compression method for Vectorcardiogram signal using hybrid autoencoder model
.png){kind=icon}
PRE
AI
DOI：10.1016/j.bspc.2025.108796.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En
Efficient compression of Vectorcardiogram (VCG) signals has become increasingly critical due to the rising demand for precise cardiac monitoring and diagnosis, coupled with storage and bandwidth limitations. This paper introduces a novel hybrid autoencoder model designed for compressing 3-D VCG signals. The preprocessing pipeline includes applying a low-pass Butterworth filter and Standardization. The proposed model architecture comprises three convolutional layers with 32, 64, and 128 filters, followed by three Long Short-Term Memory (LSTM) layers with 128, 64 and 16 units to capture long-term dependencies in the signal. The decoder mirrors the encoder using up-sampling and convolutional transpose layers to reconstruct the signal. Experimental results demonstrate that the model significantly surpasses traditional methods, such as Discrete Karhunen-Loève Transform (KLT), Discrete Cosine Transform (DCT), Discrete Wavelet Transform (DWT), and other baseline autoencoder architectures evaluated on the PTB Diagnostic ECG Database. At high compression ratio of 50, the method achieves an average fidelity of 98.81%, percentage root-mean-square difference (PRD) of 7.93%, peak signal-to-noise ratio (PSNR) of 39.42 dB, and quality score (QS) of 6.31%-1. These findings highlight the effectiveness of autoencoder in compressing the VCG signals, while maintaining high reconstruction quality, providing a promising solution for advanced cardiac monitoring systems.
Journal
IF:
4.9
Papers:
9.8K
Citations:
2.4W
Organization
No organization information available