Biomedical Signal Processing and Control 27 (2016) 134–144 Contents lists available at ScienceDirect Biomedical Signal Processing and Control jo ur nal homep age: www.elsevier.com/locate/bspc A revised scheme for real time ECG Signal denoising based on recursive filtering S. Cuomo a,∗ , G. De Pietro b , R. Farina b , A. Galletti c , G. Sannino b a Department of Mathematics and Applications, University of Naples Federico II, Via Cinthia, Naples 80126, Italy b Institute for High-Performance Computing and Networking, National Research Council (ICAR-CNR), Via Pietro Castellino 111, Naples 80131, Italy c University of Naples Parthenope, Department of Science and Technology, Naples, Italy a r t i c l e i n f o Article history: Received 22 July 2015 Received in revised form 10 December 2015 Accepted 19 February 2016 Keywords: Electrocardiogram Real time signal denoising Recursive filters Rational polynomial approximation Boundary conditions a b s t r a c t In many healthcare applications, artifacts mask or corrupt important features of Electrocardiogram (ECG) signals. In this paper we describe a revised scheme for ECG signal denoising based on a recursive filtering methodology. We suggest a suitable class of kernel functions in order to remove artifacts in the ECG signal, starting from noise frequencies in the Fourier domain. Our approach does not require high computational requirements and this feature offers the possibility of an implementation of the scheme directly on mobile computing devices. The proposed scheme allows local denoising and hence a real time visualization of the signal by means of a strategy based on boundary conditions. Experiments on real datasets have been carried out in order to test, in terms of computation and accuracy, the proposed algorithm. Finally, comparative results with other well-known denoising methods are shown. © 2016 Elsevier Ltd. All rights reserved. 1. Introduction The Electrocardiogram (ECG) signal is one of the most important tools for the detection of cardiovascular diseases. It is a graphi- cal representation of cardiac activity and it provides significant information about the functional conditions of the heart and the circulatory system. Due to the importance of the ECG signal and the related heart rate variability measurements (e.g. [1]) in the diagnosis of serious episodes, like myocardial ischemia or atrial fibrillation, the denoising signal process has become a very significant task in the medical and engineering communities [2]. Generally, the recorded signal could be corrupted by various kinds of noise signal, as discussed in [3], namely artifacts, of which the most common are well highlighted in [4]. They include power line interference, electrode contact noise, motion artifacts, muscle contraction interference, baseline drift, and some instrumental noise generated by the electronic ECG monitoring device, all of which can corrupt the ECG and lead to a wrong diagnosis. Ref. [5] studied the effects of baseline drift and power line interference on ∗ Corresponding author. E-mail addresses: salvatore.cuomo@unina.it (S. Cuomo), giuseppe.depietro@na.icar.cnr.it (G. De Pietro), raffaele.farina@na.icar.cnr.it (R. Farina), ardelio.galletti@uniparthenope.it (A. Galletti), giovanna.sannino@na.icar.cnr.it (G. Sannino). the ability of the technique to detect the morphology of some sig- nal quality features. Several methods have been applied to model and to denoise the ECG signal, such as band pass filters (e.g. [6]), adaptive filters (e.g. [7]), the ensemble averaging technique (e.g. [8]), extended Kalman filters (e.g. [9]), Wiener filtering (e.g. [10]), Empirical Mode Decomposition (e.g. [11]), and wavelet denoising (e.g. [12]). For other methods see [13]. Although some of these methods have demonstrated a good performance in terms of Signal to Noise Ratio (SNR), they can be sensitive to varying parameters. Moreover, in ECG filtering, a crucial problem is the preservation of the shape, which is achieved by several algorithms, for example by non-local means filtering, as in [14]. These methodologies belong to the off-line methods, where the electrical data are first recorded and then a denoising algorithm is applied to improve the quality of the acquired signal. This approach represents a strong limitation in several contexts, for example in healthcare systems that are under- going a transformation, as documented in [15]. Moreover, when the ECG recordings are provided by wearable sensors, that transmit data to a mobile device using wired or wireless communication, it is necessary to display and to process healthcare and biomedical information in a real time scenario. For this applicative health care context, new approaches based on novel algorithms and smart soft- ware frameworks have been widely explored to improve electronic health care (e-health) services, especially for the acquisition and filtering of biomedical signals [16–18]. In real time signal processing, two new aspects are required: the computational efficiency of the http://dx.doi.org/10.1016/j.bspc.2016.02.007 1746-8094/© 2016 Elsevier Ltd. All rights reserved.