Removal of Muscular and Artefacts Noise from the ECG by a Neural Network

The following work presents a system of cancellation muscular and artefacts noise in biomedical signals with a multilayer structure of neural networks. In this study in particular the signal has been analyzed from electrocardiogram (ECG). This system consists in a simple structure similar to the neuronal network MADALINE (Multiple ADAptive LINear Element), which is used like a structure. The proposed system is a growed artificial neuronal network which allows to optimize, the number of nodes of the hidden layer and the matrixes of coefficients. The coefficients matrix are optimized using the algorithm of simultaneous perturbation which requires a smaller computer complexity than the required one by the backpropagation algorithm. The comparison between the different typical methods (filter FIR, biorthogonal wavelet 6,8, filtered adaptive LMS) and the system based on neural multilayer networks proposed, is obtained calculating the cross correlation between the input signal to the system and the exit, in addition for the calculation of the SIR (relation parameter signal interference). The comparison shows that the neural networks method is able to better preserve the signal waveform at system output with an improved noise reduction in comparison with traditional techniques. With this method it is possible to eliminate white, artefacts and muscular noise. The rest of filter systems give worse results.