Latency variable source separation for heart rate detection in low-quality ECG signals

Monitoring of the heart rate can provide vital clinical information, but can, in specific situations, be complicated due to the low signal to noise ratio (SNR) of the available physiological signals. Several methods to enhance the SNR are known from literature, e.g. wavelet-based enhancement methods, but most of these methods require a priori information on the recorded signals and are only applicable in specific situations. In this paper a generic method is presented that uses latency variable source separation (LVSS) to derive a matched filter for enhancement of electrocardiogram (ECG) signals. Besides its use on ECG signals, the LVSS method has the potential capability to enhance any kind of (quasi-) periodical signal. The LVSS method is evaluated by comparing its performance in SNR enhancement to the performance of a wavelet-based enhancement method. This performance demonstrates that for low-SNR ECG signals, the LVSS method outperforms the wavelet-based method.