Electrocardiogram Signal Processing Using Joint Time-Frequency Tools

Abstract

Electrocardiogram (ECG), is a non-invasive technique which is used as a primary tool to diagnose cardiovascular diseases. ECG is non-stationary signal, so it cannot give good time or frequency resolution, as it is well established by the research communities that the Fourier transform is not the most viable tool for the time-frequency signal analysis and processing. So joint time-frequency analysis(JTFA) tools are introduced to get more accurate results in the non-stationary signal environment. JTFA Techniques like Short-time Fourier transform, Wigner-Ville distribution, Wavelet transform, and S-transform are discussed and contrasted to overcome the draw- backs of Fourier transform. Different parameters that make S-transform better than other transforms have been discussed. We will exploit the advantages of the S-transform to denoise the ECG signal, and for isolating the QRS complexes in the time- frequency domain. Concept of Shannon energy is introduced and applied, instead of classical squared energy to get significant amplitude changes of QRS and to emphasis medium and low QRS beats. The work presented in this thesis mainly focuses on to investigate application of JTFA tool on ECG signal. Qualitative measures like Root Mean Square Error(RMSE), Signal to Noise Ratio(SNR), Sensitivity, Positive Predictivity etc. are used for validating the results obtained for denoising and QRS detection. ECG data is taken from the standard database of MIT-BIH arrhythmia database. A modified variant of S-transform in fractional frequency domain is also proposed for biomedical signal processing. The main objective is to predict JTFA tools valuable in the biomedical field so that future developments in this field can come up.