Please use this identifier to cite or link to this item:
Title: Electrocardiogram Signal Processing and Classification
Authors: Kaur, Amandeep
Kumar, Sanjay (Guide)
Keywords: ELECTROCARDIOGRAM;Wavelet transform;Wavelet packet analysis;Adaptive boosting
Issue Date: 7-Sep-2017
Abstract: The purpose of the reported work is to provide unified introduction to the principles and applications of wavelet transform (WT) in the biomedical ECG signal. WT has proven to be powerful tool for analysis of non-stationary (time-varying) signals by providing simultaneous time-frequency resolution. As it is evident from literature that fourier transform (FT) does not reflect evolution over time of the spectrum and thus it contains no local information. WT has overcomes the drawbacks of FT by representing signals (time-dependent signals) in the phase-space (time-frequency plane), with local time-frequency resolution. Thus, WT provides an excellent time-resolution of high-frequency components and a frequency (scale) resolution of low-frequency components. There exists various wavelet based techniques such as continuous wavelet transform (CWT), discrete wavelet transform (DWT), wavelet packet analysis (WPA) as well as Mallet filtering scheme and algorithm for the DWT based calculations. In the proposed work, DWT have been thoroughly studied and applied for the analysis of biomedical ECG signal. The raw ECG data is obtained from MIT-BIH database that contains both normal as well as abnormal subjects. The filtering operation is performed on raw data for removal of noise present in the obtained ECG. Further, various artifacts are removed by denoising procedure which utilized WT techniques. Thus an appropriate mother wavelet and efficient thresholding techniques and methods will be required to obtain clean ECG free from all noises and commotions. After denoising, feature extraction is done so as to transform existing features into a lower dimensional space. The various features like P-QRS-T wave peaks, QRS, PR, RR intervals and ST segment are extracted from ECG signal. Finally, classification of ECG is performed using support vector machines (SVM), adaptive boosting (AdaBoost), random forest, neural network (NN) and decision tree. The performance matrix of classification such as accuracy, positive predicitivity and sensitivity are obtained.
Description: Master of Engineering -ECE
Appears in Collections:Masters Theses@ECED

Files in This Item:
File Description SizeFormat 
4837.pdf21.99 MBAdobe PDFView/Open    Request a copy

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.