Algorithmic Research On Non-orthogonal Wavelet Transform In Ecg Signal Processing

Heart diseases nowadays have become much more frequently happened thus arose our keen attentions. The doctors need more detailed information of which ECG signals can provide, at the same time, patients call for 24hour supervision in case of emergencies. The above mentioned requirements together adds more difficulty to ECG signal processing, which contains signal denoise, detection and compression.Wavelet Transform (WT) can provide detailed local information of signals, thanks to this inner property, plenty of researches have been carried out using WT in ECG signal processing. Based on the reseach of past algorithms,an improved non-orthogonal WT has been used. The main contributions of this thesis appear as follows:(1) Choose the Mexican-hat wavelet to be the mother wavelet function .The Mexican-hat wavelet has any-order orthogonality, this gives good frequency domain performance,and helps improve the reconstruction accuracy of ECG signal.(2) A new method using non-orthogonal based on stationary wavelet transform has been applied in ECG signal de-noising. Because of its invariability in the time domain, it gives better results comparing past mentioned DWT algorithms in ECG signal de-noising.(3) In QRS detection, slope law~a new judgment law has been applied to false max-min pair elimination .The idea roots from QRS inner property, so besides the excellent performance in max-min pair elimination, it also suggests easier algorithm implementation. The bring in of alterable threshold can greatly reduce the bad judgment occurrence and save computer time as well.(4) A new class of non-orthogonal based binary wavelet transform has been used in ECG signal compression. The coefficients of the DWT are calculated such that the square of the difference between the original signal and the reconstructed one is minimum in least mean square sense. Analysis show that this technique outperforms other traditional DWT technique by yielding the smallest PRD under given compression ratio (CR≤10) with the significant diagnosing features well preserved.