International Journal of Electrical and Computer Engineering (IJECE) Vol. 6, No. 5, October 2016, pp. 1 – 8 ISSN: 2088-8708 1 Empirical Mode Decomposition (EMD) Based Denoising Method for Heart Sound Signal and Its Performance Analysis Amy H. Salman, Nur Ahmadi, Richard Mengko, Armein Z. R. Langi, and Tati L. R. Mengko School of Electrical Engineering and Informatics, Institut Teknologi Bandung Article Info Article history: Received May 26, 2016 Revised June 30, 2016 Accepted August 12, 2016 Keyword: Heart Sound Denoising Empirical Mode Decomposition ABSTRACT In this paper, a denoising method for heart sound signal based on empirical mode decompo- sition (EMD) is proposed. To evaluate the performance of the proposed method, extensive simulations are performed using synthetic normal and abnormal heart sound data corrupted with white, colored, exponential and alpha-stable noise under different SNR input values. The performance is evaluated in terms of signal-to-noise ratio (SNR), root mean square er- ror (RMSE), and percent root mean square difference (PRD), and compared with wavelet transform (WT) and total variation (TV) denoising methods. The simulation results show that the proposed method outperforms two other methods in removing three types of noises. Copyright c  2016 Institute of Advanced Engineering and Science. All rights reserved. Corresponding Author: Amy H. Salman School of Electrical Engineering and Informatics Bandung Institute of Technology Jl. Ganesha No 10, Bandung 40132, Indonesia Email: amy@stei.itb.ac.id 1. INTRODUCTION Cardiovascular disease (CVD) has long been the leading cause of death throughout the world with an estimate of 17.3 million people died in 2008 and is predicted to reach 23.3 million in 2030 [1]. According to WHO report, more than three quarters of the death takes place in low- and middle-income countries [2]. A low-cost and non-invasive diagnosis system based on heart sound can be used to minimize the risk of patients going into severe condition and reduce the financial burden through an early accurate diagnosis followed by appropriate treatment. This electronic auscultation technique utilizes advanced signal processing with fast computation capability, thanks to the advancement of computer technology. However, to produce an accurate diagnosis result is not an easy task since, in practice, heart sound signal is always contaminated with noise and interference from various sources such as background noise, power interference, breathing or lung sounds, and skin movements in the surrounding environment. Thus, signal denosing method is of paramount importance to remove all these unwanted noise. A poor signal denoising method can lead to catastrophic result. The most widely used method for denoising heart sound signal is based on wavelet transform (WT) [3–5], a powerful signal analysis tool with the ability to represent a signal simultaneously in the time and frequency. Despite the fact that the wavelet based denoising method has been proven to be able to provide good denoising performance, however, it suffers from several limitations. It requires predefined basis function selection (from too many choices) suited to signal under consideration, which limits the flexibility of the method. In addition, the decomposition level and thresholding technique of wavelet denosing also need to be carefully considered. Failing to choose the right de- composition level and thresholding technique will result in bad denoising performance. Varghees and Ramachandran employed another alternative method based on Total Variation (TV) [6]. TV method has been mostly used for image denoising due to its great benefit of preserving and enhancing important features such as edge in images. Even though it can be used for denoising 1D signal, nevertheless, there are very few literatures exploiting TV method for deonising heart sound. The highly non-stationary property of heart sound signal is not suitable to the nature of TV method which performs best on piecewise constant signals [7]. In addition, Figueiredo et. al. mentioned that the performance of TV Journal Homepage: http://iaesjournal.com/online/index.php/IJECE