Title :
Wavelet match filtering and neural network based QRS detection
Author :
Hendija, Dubravko
Author_Institution :
Fakultet Elektrotehnike i Racunarstva, Zagreb, Croatia
Abstract :
This paper presents combination of wavelet match filtering and neural network approach in QRS detection. In development, a particular emphasis is put on low signal-to-noise ratio and low computational complexity. Morlet wavelet is used for artifact removal and MLP is then used for QRS classification. Testing on MIT/BIH arrhythmia database, with added artifacts, shows above 90% accuracy in QRS detection in worst case scenario.
Keywords :
computational complexity; electrocardiography; medical signal detection; neural nets; wavelet transforms; MIT/BIH arrhythmia database; Morlet wavelet; QRS classification; QRS detection; artifact removal; computational complexity; neural network; signal-to-noise ratio; wavelet match filtering; Artificial neural networks; Databases; Electrocardiography; Filtering; Neurons; Noise; Wavelet transforms;
Conference_Titel :
MIPRO, 2011 Proceedings of the 34th International Convention
Conference_Location :
Opatija
Print_ISBN :
978-1-4577-0996-8
