engleski

Comparison of the SVM classification results between original and DWT denoised respiratory signals considering to the transients noise

Asthma and COPD are the most common breathing diseases nowadays. Analysis and processing of breathing records are important diagnosis tools. This paper compares efficiency results of SVM classification of two classes: first the breathing noise and second the pause of signal samples recorded on subjects in real-life clinical conditions. In these conditions there is an appearance of noise and short impulse signals (transients) which are in this paper reduced by using DWT and different thresholding techniques to see if the denoised signal samples give better validation results. Classification results show that the best results are obtained when energy, Renyi entropy and standard deviation are taken as features for SVM classification. Testing of data revealed that original signal samples give better results of accuracy than the denoised signal samples.

phonopneumogram, wavelet, classification, accuracy

nije evidentirano