Algorithm study on ECG diagnosis based on wavelet packet decomposition combined with DPFNN

An Algorithm study on ECG diagnosis based on wavelet packet decomposition combined with DPFNN (double parallel feedforward neural net) is proposed. Signal features are identified through the ECG data analysis. The frequency bands of ECG signals including identified feature are extracted by wavelet packed decomposition. In order to overcome local minima and speedup the convergence of BP, a novel DPFNN net is proposed. The experimental results demonstrate that the algorithm can speed up the convergence rate of 2–3 times, and can extract signal features effectively and identify signal exactly .