Sigmoidal radial basis function ANN for QRS complex detection

This work proposes a QRS complex detector which works on high amplitude. An optimized nonlinear adaptive whitening filter based on sigmoidal radial basis function (SRBF) is used to cleverly accentuate the QRS complexes and suppress the noises present in the ECG signal. An M-estimator based objective function and a new contextually based centering technique are used to control the behavior of the whitening filter so that it totally suppresses low frequencies segments and preserves almost unchanged the high frequencies ones. The residual issued from the whitening filter is then fed to a matched filter to enhance the signal-to-noise ratio (SNR) which is yet an important step that allows achieving good detection performance. Additional nonlinear transformations are performed before the decision stage namely squaring and moving average filtering. The validation of the detector is performed over the whole MIT/BIH database files. The algorithm has produced an average detection error rate of 0.28%, sensitivity of 99.82% and specificity of 99.91%.