The Research Of Intelligent Recognition Method To Abnormal ECG Wave

Heart disease has been threatening to people's life seriously. Electrocardiogram (ECG) is widely used in heart disease diagnosis. However, the abnormal ECG waves greatly increase difficulties in recognizing ECG signals by computer automatically and affect its recognition accuracy. An intelligent recognition method to detect abnormal ECG wave is presented in this thesis. The abnormal wave is here divided into two types:first type of abnormal wave and second type of abnormal wave. The first type of abnormal wave includes frequency interference abnormal wave and dynamic baseline drift. For the first type of abnormal wave we use denoising strategy which is based on the wavelet transform as core principle, decreasing noise to power frequency abnormal wave by Multi-scale wavelet threshold denoising method, getting low-scale high-frequency signals through wavelet transform. Then different thresholds are used for different level of signals. For dynamic baseline driftting, high-scale approximation corresponding frequency band signals are extracted by wavelet transform, and then signals are reconstructed. For the second type of abnormal wave, five kinds of features are used as identifying strategy including characteristics of standard deviation, characteristics of singular points, characteristics of ST-segment fitting errors, FFT fitting errors and characteristics of signal peak points. Experimental results show that the strategy can effectively detect the second type of ECG abnormal wave fragments.