Wavelet Based Distortion Measurement and Enhancement of ECG signal Md. Abdul Awal 1 , Mohiuddin Ahmad 2,3 1,2 Department of Biomedical Engineering 3 Department of Electrical and Electronic Engineering Khulna University of Engineering & Technology Khulna-9203, Bangladesh Email: mohiuddin0ahmad@gmail.com I. Daut 4 , E. C. Mid 5 , M. A. Rashid 6 4,5,6 School of Electrical Systems Engineering University Malaysia Perlis (UNiMAP) Perlis, Malaysia E-mail: abdurrashid@unimap.edu.my Abstract—This paper describes the research carried out to eliminate the noise found in ECG signal and cardiac rhythm using Coiflet wavelet since its scaling function is closely related to the shape of ECG and suited for denoising for many applications. Different adaptive and soft threshold functions are used to enhance the ECG signal. The output SNR shows an outstanding performance of 6.2 dB as compared with other methods. In addition to SNR improvement and different statistical tools, a new time-scale qualitative distortion measure named scalogram difference factor is also proposed. Keywords-ECG, Coiflet wavelet, denoising, scalogram difference factor, SNR improvement. I. INTRODUCTION The electrocardiogram (ECG) is the recording of the cardiac activity obtained by a noninvasive technique which provides useful information for the detection, diagnosis and treatment of cardiac diseases .However, for the sampled ECG signal, it is inevitable to mix wanted signal with varied noises such as white noise, pink noise, baseline wander, muscle noise, motion artifact and so on, which in varying degrees causes misjudgment and omission of conventional ECG identification for the ECG features extraction and reduces the diagnostic accuracy. In addition, with the recent telemedicine applications involving transmission and storage of ECG, noise also appears due to poor channel conditions. A noisy ECG may hinder the physician’s correct evaluations on patients. Therefore denoising and preprocessing of ECG signal becomes an exclusive requirement. Among several approaches reported so far to address ECG enhancement adaptive filter architecture is commonly used [1]. Statistical techniques such as principal component analysis, independent component analysis, and neural networks [2], have also been used to extract a noise-free signal from the noisy ECG. Over the past several years, methods based on the wavelet transform (WT) have also received a great deal of attention for denoising of signals that possess multi-resolution characteristics. Baseline estimation using cubic spline, baseline construction by linearly interpolating between pre-known isoelectric levels estimated from PR intervals [3], linear filtering, and the use of wavelet packets are major approaches in this field. To remove the noise level in the signal using wavelets, it must be selected from those similar to ECG waveforms, like the ones developed by Daubechies, Coiflets etc. [4]. In this research, Coifleties wavelet function has been selected for denoising and the basic aim is to improve the quality of the ECG signal using various wavelets based adaptive threshold techniques. Furthermore, different statistical tools like Percent Root Difference (PRD), Mean Square Error (MSE), Normalized MSE (NMSE) and Root Mean square Error (RMSE) as well as SNR improvement and Scalogram difference factor (SDF) is proposed to evaluate the performance. This paper is organized as follows. Section II provides theoretical background on the definition of wavelet transform. Section III focuses on the different adaptive threshold schemes. In Section IV, brief characteristics of noises and their implementations are presented. Section V describes the quantitative and qualitative evaluation of the techniques. Finally, discussion and concluding remarks are provided in Section VI. II. WAVELET TRANSFORM A. Continuous Wavelet Transform The wavelet is a scaled and shifted copies of the main pattern, so called the “mother wavelet” are known as wavelets. In a nutshell, the continuous wavelet (CWT) is nothing but a set of the inner product of the observed signal f(t) with the shifted and scaled mother wavelets is given by ) ( 1 ) ( a, a t a t τ ϕ ϕ τ - = (1) where τ and a (>0) represent the time shift and scale variable respectively. dt a t t f a a WT t f f a ∫ - = = ∗ ) ( ) ( 1 ) , ( ), ( , τ ϕ τ ϕ τ (2) 2012 International Conference on Biomedical Engineering (ICoBE),Penang,Malaysia,27-28 February 2012 978-4577-1991-2/12/$26.00 ©2011 IEEE