Vector Quantization for ECG Beats Classification

Reducing the feature dimensionality can improve the computational efficiency of electrocardiogram (ECG) beats classification system. In the long term ECG classification task, vector quantization has demonstrated its advantage in both dimensionality reduction and accuracy increase, but the existing vector quantization methods are not capable of representing the difference of each waveform among ECG beats. To make vector quantization available for ECG beats classification, in this paper, we propose a strategy that aligns each wave of all beats, and then build a dictionary corresponding to each wave segment. Thus vector quantization can distinguish each waveform of different beats. We compare our method with the popular beats features such as sampling point feature, fast Fourier transform feature, and discrete wavelet transform feature. The classification results show that our feature has high accuracy and is capable of reducing computational complexity of beats classification system, which demonstrate that the proposed method can provide an effective vector quantization feature for beats classification.