Self-training dictionary based approximated l0 norm constraint reconstruction for compressed ECG

Compressed sensing provides a sub-Nyquist sampling rate but requires a sparse structure of the signal. Electrocardiogram (ECG) signal expresses no sparse structure in time domain. This study aims to improve the sparse expression based on a self-training dictionary scheme (STDS), to obtain a self-training dictionary. Furthermore, an approximated l0 norm constraint method (AL0CM) is designed for sparse expression by using an accelerated gradient descent method. The reconstruction of the compressed data could be applied by combining the sampling matrix, dictionary, and sparse expression. The ECG experiments are conducted to confirm the superiority of the proposed STDS-AL0CM framework in terms of the recovered signal-to-noise ratio (RSNR), percent norm difference (PND), and running time.