Research On Noise Elimination Of ECG And Its Denoising Evaluation

Noise such as baseline wander in ECG (Electrocardiogram) signals is a long termproblem to interfere the quality of ECG signal. The methods of denoising were emergedendlessly. According to the characteristics of medical diagnosis information in ECG and therestricted requirements on signal authenticity, great efforts had been made in this paper toeliminate the baseline wander and to get a signal with clear characteristic points. The resultsof that were used in following diagnosis. All the works were under the condition ofminimizing signal morph distortion. Morphology was selected to reduce the baseline wanderin any shape of ECG waveforms. It will cause the least distortion. Its structure element wasexpanded to two dimensions instead of the traditional one dimension structure element. Fromthe result, two-dimension structure element gained a better control in denoising procession.Meanwhile, the morphology method with two-dimension structure element was describedin mathematical formulas by derivation. It simplified the computing of the morphology inprocessing signals. The formulas that derived from the vary definitions of morphology holdtheir universal property in baseline extraction of other signals. Beside this, a brief designationprinciple of structure element was proposed as a rule; and the two-dimension structureelement had been proved by an instance to work more effectively in distortion control than theone dimension structure element. The denoising of two-dimension structure element could bemodified and gain better quality by employing some signal compensations.Till now, there is not a widely accepted evaluation standard for ECG denoising andespecially for distortion has been presented. A system with quantized indicators was proposedfor the evaluation of the noise elimination and the distortion control in the paper. The systemwas according to the engender principium and clinical meanings of ECG. This evaluationsystem covers all the characteristic points and their intervals, potentials and some items whichinfluence the clinical diagnosis. Through computing, it gave a numerical value for impersonalevaluation denoising and distortion control ability of those strategies.At last, in the experiments,8samples of the MIT/BIH Arrhythmia Database wereselected, correspondingly to8typical sorts of ECG waveforms. After adding artificial wander,those data were processed by the morphology based on two-dimension structure element andone dimension structure element. Same works were done by the wavelet, and the median.Their denoising and distortion control effectiveness were evaluated by the evaluation systemof this paper. From the result, two-dimension structure element method is slightly less restfulthan that of one dimension structure element methods in denoising. But it is the mosteffectively control of distortion.