Denoising of ECG Signal using UFIR Smoothing with Notch Filter

Abstract

The electrical activity of the heart is test with an electrocardiogram (ECG). The fundamental information for the taking decision about various types of heart diseases identified by electrocardiogram. There have been numerous attempts over decades to extract the characteristics of the heartbeat through ECG records with high accuracy and efficiency using a variety of strategies and techniques. In this paper a novel scheme is acquainted, the problem is solved by isolated time space using q-lag unbiased finite impulse response (UFIR), then the received time changing of optimal average horizon for the shape of the ECG signal. A complete statistical analysis is furnished by normalized histogram and statistical classifiers, P wave features extraction based on the detected fiducial points is deliberated. In this concept by utilizing QRS detection, morphological top-bottom hat transformation and notch filters is ameliorated PSNR and latency constraints, furnishes high accuracy and reduced elapsed time.

Introduction

I. INTRODUCTION

The electrocardiogram acquainted to analysis the rhythmical throbbing of the arteries as blood is propelled through them. An electrocardiogram records the electrical signal in your heart. It is a simple and painless test employed to quick diagnose heart problems and monitor your heart health. The electrical signals from the heart to check various heart conditions, but it is susceptible to noises. ECG denoising is a major pre-processing step which attenuates the noises and emphasize the normal waves in ECG signals. In particular several algorithms have been developed to analyze and capture the reliability properties and rhythm variables in ECG signals, based on the clustering of syntactic features to detect noise and extract information about atrial behavior. Learned by P through ECG signals, QRS, and T waves, appropriate methods of ECG signals denominating and feature extraction are utilized. However, reaching accurate results is still challenging due to the data collection equipment.

Here we introduced an ameliorated unbiased finite impulse response(UFIR),impartial sensitivity only for odd degree polynomials correct lag q should be taken from other points for even degree polynomials. P shift theory, q=-p>0, unbiased finite impulse response (UFIR) filtering, Savitsky-Golay is considered sensitive in particular to odd order polynomials, to provide the best denoting effect, this approach must be set individually for each polynomial indicating the correct lag q and not necessarily at the midpoint.

Furthermore, it provides smoothing filtering by p<0, filtering by p=0 and predictive filtering by p>0.The Savitsky-Golay filter was recently modified to be optimal in the minimum mean square error (MSE) sense, the modification is equivalent to a proper UFIR filter, which produces the maximum probability estimate because these two solutions require information about the noise that is not well studied in the ECG signals. An optimal q lag state space UFIR smoothing algorithm is proposed for the ECG signals denoising, artifact removal and stable features evaluation using different classifiers under unknown noises of the ECG signals. It provides high accuracy pattern classification for ECG signals and ameliorated PSNR and better RMSE values, furnishes feature extraction and reduces time elapsing.

For the rest of the paper is organized as follows, ECG Signal Data Base and Model are described in section ?, followed by the Features Extraction in State Space using an UFIR Smoother in section ?, in section ? discusses Simulation Results and finally the conclusion of the work is discussed in section ?.

II. ECG SIGNAL DATABASE AND MODEL

MIT-BIH arrhythmia database is publicly available dataset which provides standard investigation material for the detection of heart arrhythmia. It is used for purpose of fundamental research and medical device development on cardiac rhythm and related diseases. This work employs the MIT-BIH Arrhythmia Benchmark, which contains many records taken from different databases, such as the MIT-BIH Arrhythmia (MITDB). MITDB holds 48 records with simple and extraordinary rhythms taken from 47 subjects. Records were sampled up to 360 Hz per lead with 11-bit resolution in the 10mV range. This database provides records in two leads, the most common being MLII (Modified Lead II). Other leads, such as V1, V5, etc. are also used. An important issue is choosing a bottle that most clearly reflects the ECG signal morphology.

A. ECG Signal Model in Discrete-Time State-Space

To provide effective denoting and extraction of features, in this subsection we will model the ECG signal in isolated-time state-space. We refer to the ECG signal‌ on the horizon [m, n] at n points from m = n - N + 1to n, where n is the discrete time indicator. Inherent ECG noise is not yet well understood and its misinterpretation can cause assessment errors. Therefore, we assume that the underlying process in each ECG pulse is time-constant and decisive. We assume that the scalar dimensions of the ECG signal are provided in the presence of an unknown distribution (not necessarily Gaussian) and zero mean noise with a coverage. According to such estimates, we represent the ECG signal at discrete-time status-space with the following status and observation equations, respectively.

The QRS complex between Qp and Sp is processed with Nmin to follow a rapid excursion around Rp from Sp to Sint, Horizon Napt simply increases from Nmin to Nopt, the horizon eventually becomes Nopt above Sint. Accordingly, custom UFIR smoothing is provided as

Here T represents the heartbeat length. If Napt is provided, we can design a UFIR smoothing algorithm using subsequent iterations, which will reduce the computational load.

III. FEATURES EXTRACTION OF STATE SPACE USING UFIR SMOOTHER

The ECG signal in state space using an unbiased finite impulse response (UFIR) smoothing features extraction are provided in five stages as shown in the figure 1, which consists of different stages 1) Detrend, 2) QRS complex detection, 3) Segmentation, 4) UFIR smoothing, 5) Windowing of ECG signals, 6) Fiducial Detection, 7) Notch filter, 8) Interference removal by top-bottom hat transformation and 9) Interference removal.

1) Detrending: ECG demonstrated a detrending method to eliminate the level of RR interval fluctuations in data b and tested its properties. The detrending approach sensitivity predicts the trend of a given signal based on pre-formulation. The process of removing the baseline wandering in the ECG signal by the detrending method begins with setting the cutoff frequency by determining the regulation parameter, unlike the FIR and IIR digital filters, the total segment of the input signal is considered during the detrending operation and no time is obtained to get processing signal.

2) QRS Complex Detection: QRS detection is a basic step in determining the heart rate for subsequent rhythm classification, so the high QRS detection rate method is the most important component of the ECG analysis algorithm. The QRS-complex was discovered using citations from the arrhythmia MIT-BIH database following the procedure proposed by Pan Tompkins. Note that the highest number of citations identified the QRS complex with a probability of 93.4%.

3) Segmentation: The QRS-complex is localized, the point closest to the R-peak in each heart rate. Next, by taking 100 samples to the left and 200 samples to the right, a window is created to describe the heart rate as shown in Figure 1. Does not cover all points of interest (P-wave, QRS-complex and T-wave), its width increased. The segmentation process is performed heuristically with the aim of analyzing the morphological waves.

4)  UFIR Smoothing: We will progress an effective computational algorithm for the extraction of ECG signal features. For this, we first localize specific points on the ECG heart rate pulse and then calculate the corresponding amplitude, duration, and angle. Unlike the developed methods, this algorithm is based on the UFIR Smoothing flutter used with p-shift and l = 2 and p <0. Nopt = 21 suites were found for the softer parts of the isolated ECG signal and the data used as Nmin = 3 feet QRS complex. Note that Nopt and Nmin must be specified for the measured ECG signals. The Nopt and Napt described for the database used apply beyond the Horizon Nopt = 21 QRS complex. To avoid large bias errors, Napt is specified and applied to all EGC signals. The UFIR filtering provided is performed using smoothing with a lag q.

5) Windowing of ECG Signals: The UFIR smoother delivers denoting and evaluation of the three states of the ECG signal as shown in Figure 2, for the first state consists of denoised ECG signals, the second state is time derivative of the denoised signal and the third state is the second derivative of the denoised signal. Using information about ECG signal status, the R-peak, QRSmax and QRSmin are determined and a window is applied to cover the QRS complex. P-Wave Detection is provided starting at Q and ending with the heart rate. Here, in the second case a window is applied to cover the Pon, P-peak, Poff points determined by Pmax and Pmin. Similarly, the T-wave is detected, in which case the Ton and Tof are covered by a window created for Tmax and Tmin.

7) Notch Filter: Notch filter is also called as band stop filter or band reject filter. These filters attenuate signals within a specific frequency band called the stop band frequency range, and pass signals upper and lower this band. For example, if a notch filter has a stop band frequency of 1000mhz to 1050mhz, it sends signals from DC to 1000mhz and more than 1050mhz. It only blocks signals from 1000 MHz to 1050mhz. The optimal response to any notch filter is a completely flat response within the usable range except for the notch frequency. Here it drops very fast by providing a high level of attenuation that can remove the unwanted signal.

8) Interference Removal by top-bottom hat Transformation: Morphological signal processing involves a wide collection of theoretical concepts and mathematical tools for signal analysis, non-linear signal operators, design techniques and application systems related to mathematical operations. The morphological operators of opening and closing are very easy, and the morphological top-hot transformation that arises from these operators has confirmed to be powerful tools and has been utilize in a variety of applications, giving accomplished results in terms of noise reduction, edge detection and object identity. Top-hat transformation (STH) is acquired by subtracting the signal opening from the original signal, while the bottom-hat transition (SBH) is acquired by removing the original image from the signal closed signal.

IV. SIMULATION RESULTS

As reported by simulation results are extracted by the given input signal is evolved by using MAT Lab software. Here an input ECG signal is taken in fig.4(a) a noise signal added to the ECG signal fig.4(b) then the resultant of ECG with noise signal fig.4(c) is evolved. After filtering and removal of artifact Measurement residuals produced by the UFIR smoother (q-lag1 and q-lag2). The first-time derivative signal is produced by using windowing of the ECG signal. By conducting the accurate features extraction of the signals, the P wave features is extracted and produces normalized histogram. The interference is removed by using top-bottom hat transformation and notch filter attenuate the noise signal and finally gets the denoised ECG signal.

Conclusion

This method proposed a novel technique which remove the measurement of noise and extract concurrently features of ECG signal in state space using unbiased finite impulse response (UFIR). This smoother does not require the noise statistics and initial values and is thus more suitable for ECG signals, whose noise is still not well understood. This approach involves power line interference and baseline navigation of ECG signals, such as notch filtering and morphological filtering. Simulation results show that the proposed paper provides better performance of baseline wandering elimination for ECG signals and reduces the root mean square error (RMSE). The Specific advantages due to its suitability to non-stationary and non-linear time series. May be even the most complex problem that has not yet been solved is that the biomedical timeline is often recorded over long periods of time that extend over days and weeks. The state-space UFIR smoothing approach enhanced in this process for ECG signal denoising and feature extraction has demonstrated better results than existing methods.

References

[1] Sargolzaei, S., Faez, K., &Sargolzaei, A. (2008, May). Signal processing based for fetal electrocardiogram extraction. In 2008 International Conference on BioMedical Engineering and Informatics (Vol. 2, pp. 492-496). IEEE. [2] Kimura, Y., Sato, N., Sugawara, J., Velayo, C., Hoshiai, T., Nagase, S., ...&Yaegashi, N. (2012). Recent advances in fetal electrocardiography. The Open Medical Devices Journal, 4(1). [3] Fanaswala, M. (2005). Fetal Electrocardiogram Extraction: A Case-Study in Non-linear System Identification. [4] Ferrara, E. R., &Widraw, B. (1982). Fetal electrocardiogram enhancement by time-sequenced adaptive filtering. IEEE Transactions on Biomedical Engineering, (6), 458-460. [5] Kam, A., & Cohen, A. (1999, March). Detection of fetal ECG with IIR adaptive filtering and genetic algorithms. In 1999 IEEE International Conference on Acoustics, Speech, and Signal Processing. Proceedings. ICASSP99 (Cat. No. 99CH36258) (Vol. 4, pp. 2335-2338). IEEE. [6] Talha, M., Guettouche, M. A., &Bousbia-Salah, A. (2010, November). Combination of a FIR filter with a genetic algorithm for the extraction of a fetal ECG. In 2010 Conference Record of the Forty Fourth Asilomar Conference on Signals, Systems and Computers (pp. 1756-1759). IEEE. [7] Kholdi, E., Bigdeli, N., &Afshar, K. (2011). A New GA-Based Adaptive filter for fetal ECG extraction. World Academy of Science, Engineering and Technology, 54. [8] Niknazar, M., Rivet, B., &Jutten, C. (2012). Fetal ECG extraction by extended state Kalman filtering based on single-channel recordings. IEEE Transactions on Biomedical Engineering, 60(5), 1345-1352. [9] Panigrahy, D., Rakshit, M., &Sahu, P. K. (2015, May). An efficient method for fetal ECG extraction from single channel abdominal ECG. In 2015 international conference on industrial instrumentation and control (ICIC) (pp. 1083- 1088). IEEE. [10] Khamene, A., &Negahdaripour, S. (2000). A new method for the extraction of fetal ECG from the composite abdominal signal. IEEE Transactions on Biomedical Engineering, 47(4), 507-516.

Copyright

Copyright © 2022 A. Rajani, N. Pavani. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.