U2013 Proceedings of International Conference on Modelling, Identiication & Control (lCMIC) Cairo, Egypt, 31 stAug.- 2ndSept. 2013 Self Tuned SSRLS Filter for Online Tracking and Removal of Power Line Interference from Electrocardiogram Nauman Razzaq, Maryam Butt, Muhammad Salman, Rahat Ali, Ismail Sadiq, Khalid Munawar, Tahir Zaidi Abstact- Power-Line Interference (PLI) may seriously distort the ECG (Electrocardiogram) signal rendering the ECG analysis inaccurate. Various types of Notch iltering and adaptive cancellation schemes have been used to eliminate PLI interference from ECG. In this paper we propose an enhanced adaptive noise canceller (ANC) based on State Space Recursive Least Square (SSRLS) for elimination of PLI. SSRLS algorithm is a state space model-dependent recursive algorithm characterized by fast rate of convergence and hence exhibits excellent tracking performance. State-space formulation of noise model enables SSRLS ilter to be implemented in an instinctive manner in comparison to other approaches. The proposed method is very effective for online tracking of PLI without requiring a reference noise source. The proposed method can also be employed to ilter out the PLI even when corrupted ECG signal has low signal-to-noise ratio (SNR). Index Terms-Electrocardiography; Power line interference; Adaptive noise canceller; SSRLS; Frequency estimation; Discrete Fourier Transform; I. INTRODUCTION E L ECTROCARDIOGRAPHY (ECG) is a noninvasive method, commonly used for monitoring heart's electric activity to determine the cardiac health. Beside other noises present in ECG signal, the most disturbing noise is power line interference (PLI). A good PLI elimination technique is targeted to remove only PLI component, while keeping the Manuscript received on May 5, 2013. This work is supported by National Information & Communication Technologies Research and Development (lCT R&D) Fund project and National Institute of Heart Diseases NlHD), Pakistan. Nauman Razzaq, a corresponding author is with department of Electrical Engineering, College of E&ME, National University of Sciences and Technology NUST), Islamabad, Pakistan. (Phone: +92-51-9278055 ext 4280; Fax: +92-51-9278049; e-mail: nauman.razzaq@ceme. nust.edu.pk). Maryum Butt is with department of Electrical Engineering, College of E&ME, National University of Sciences and Technology, Islamabad, Pakistan. (email: maryambutt2011@ceme. nust.edu.pk). Muhammad Salman is with department of Electrical Engineering, College of E&ME, National University of Sciences and Technology, Islamabad, Pakistan. (email: salmanmasaud@ceme.nust.edu.pk). Rahat Ali is with deprtment of Electrical Engineering, College of E&ME, National University of Sciences and Technology, Islmabad, Pakistan. (email: rahat.ali@ceme.nust.edu.pk). Ismail Sadiq is with Center For Advanced Studies In Engineering, Islamabad, Pakistan. (email: ismail_sadiCLee@yahoo.com) Khalid Munawar is with department of Electrical and Computer Engineering King Abdulaziz University, Jeddah, Saudi Arabia. (email: kmunawar@kau.edu.sa). Tahir Zaidi is with with department of Electrical Engineering, College of E&ME, National University of Sciences and Technology, Islamabad, Pakistan. (email: tahirzaidi@ceme. nust.edu.pk). 339 ECG signal intact. Implementation of a ixed IIR notch ilter is the simplest and direct scheme for removal of PLI disturbance. However, this ilter cannot work well when the PLI requency drits beside having other limitations of transient response and ringing effects [1]. In such cases, adaptive notch ilter is necessary, whose response is adjusted automatically through some variable parameter according to the speciied criterion. However, an adaptive ilter requires a reference input, which is not always available in all equipment especially in battery operated portable equipment. B. Widrow uses a scheme of adaptive iltration where PLI is also provided as reference input [2]. Adaptive Sinusoidal Interference Canceller (ASIC) has been proposed by H.C. So in which requency of PLI is considered to be known [3]. S ome adaptive iltration techniques do not require a reference input, but are generally computationally expensive, which makes them unworkable for online implementation. Bazhyna et al. have proposed a solution in which they isolated the linear and non linear segments of ECG and then use the linear segment to estimate the PLI which cause time lag as well as computational overload [4]. This paper investigates a new adaptive method for elimination of PLI rom ECG signal without the requirement of auxiliary reference input. In this method a State Space Recursive Least Square (SS RLS) adaptive ilter [5] in combination with Fourier based requency estimation is implemented. The proposed method works as adaptive noise canceller by subtracting estimated PLI rom the corrupted ECG signal. II. METHODS AND MATERIAL A. Noise Modeling The discrete-time PLI may be defmed as X P /) = a sinrfL / sk+p) (1) Where k is the discrete time index, ; is sampling requency and iL is the requency of PLI, p is phase of PLI and a is amplitude ofPLI. Let us deine the PLI model in State Space and the two states of a 2n d order state space model be X l (k) = a sin(2r s k +p) (2) X 2 k) = a cos(2r � k +p) Using trigonometric identities, (2) can be rewritten as