A new QRS detector based on neural network
The application of multi-layer perceptron, artificial neural network (ANN) model, to detect the QRS complex in the ECG signal processing is presented. The target is to make the neural network analyze directly the raw ECG signal (without any preprocess) through a moving temporal window, and detect the presence of QRS complex by using only a single fixed threshold algorithm on the output. Preliminary experiment results, performed by fixing the majority of the large number of the system degrees of freedom, indicate that this new method is quite promising. This article also presents a two-stages trainning method that has revealed to be very efficient in the improvement of the learning process of the perceptron.
Unable to display preview. Download preview PDF.
- Tompkins, W. Biomedical Digital Signal Processing. Prentice-Hall 1993, pp. 250–270.Google Scholar
- Zhu K., Noakes P. and Green P. “Training Neural Netwoks for ECG feature recognition”. International Neural Network Conference (INNC'90), Vol. 1, pp. 137–140.Google Scholar
- Hilera J. and Martínez V. “Neural networks for ECG compression and classification”. 3rd International Conference on Fuzzy Logic and Neural Nets IIZUKA94, pp. 121–124.Google Scholar
- Leong P. and Jabri A. “A low power VLSI arrhythmia classifier”. IEEE Trans. Neural Networks, Vol. 6, No. 6, Nov. 1995, pp. 1435–1445.Google Scholar
- MIT/BIH Database Distribution, Massachussetts Innst. Technol., 77 Massachussetts Avenue, Room 20A-113, Cambridge, MA 02139.Google Scholar
- Reinhardt J. and Müller B. Neural Networks. An Introduction. Springer-Verlag 1991, pp. 223–228.Google Scholar
- Xue Q., Hen Y. and Tompkins W. “Neural Network based Adaptative Matched Filtering for QRS detection”. IEEE Trans. Biomed. Eng., Vol. 39, No 4 April 1992, pp. 317–339.Google Scholar