International Journal of Electrical and Computer Engineering (IJECE) Vol. 12, No. 5, October 2022, pp. 4944~4950 ISSN: 2088-8708, DOI: 10.11591/ijece.v12i5.pp4944-4950  4944 Journal homepage: http://ijece.iaescore.com Electrocardiograph signal recognition using wavelet transform based on optimized neural network Ali Talib Jawad 1 , Dalael Saad Abdul-Zahra 2 , Hassan Muwafaq Gheni 3 , Ali Najim Abdullah 4 1 Department of Medical Instrumentation Technologies Engineering, Hilla University College, Babylon, Iraq 2 Department of Medical Physics, Hilla University College, Babylon, Iraq 3 Computer Techniques Engineering Department, Al-Mustaqbal University College, Hillah, Iraq 4 Department of Medical Instrumentation Technologies Engineering, Hilla University College, Babylon, Iraq Article Info ABSTRACT Article history: Received Oct 18, 2021 Revised Jun 10, 2022 Accepted Jun 23, 2022 Due to the growing number of cardiac patients, an automatic detection that detects various heart abnormalities has been developed to relieve and share physicians’ workload. Many of the depolarization of ventricles complex waves (QRS) detection algorithms with multiple properties have recently been presented; nevertheless, real-time implementations in low-cost systems remain a challenge due to limited hardware resources. The proposed algorithm finds a solution for the delay in processing by minimizing the input vector’s dimension and, as a result, the classifier’s complexity. In this paper, the wavelet transform is employed for feature extraction. The optimized neural network is used for classification with 8-classes for the electrocardiogram (ECG) signal this data is taken from two ECG signals (ST-T and MIT-BIH database). The wavelet transform coefficients are used for the artificial neural network’s training process and optimized by using the invasive weed optimization (IWO) algorithm. The suggested system has a sensitivity of over 70%, a specificity of over 94%, a positive predictive of over 65%, a negative predictive of more than 93%, and a classification accuracy of more than 80%. The performance of the classifier improves when the number of neurons in the hidden layer is increased. Keywords: Electrocardiogram recognitions Invasive weed optimization Optimized neural networks Patterns recognition Wavelet transforms This is an open access article under the CC BY-SA license. Corresponding Author: Ali Talib Jawad Department of Medical Instrumentation Technologies Engineering, Hilla University College Babylon, Iraq Email: alitaleb54qq@gmail.com 1. INTRODUCTION As a cost-effective and non-invasive method of observing heart function. The electrocardiogram (ECG) was commonly utilized [1]. The ECG signal reflects heart functionality, which helps the cardiologist diagnose cardiac problems. Cardiovascular disorders are the leading cause of death, the ECG field has advanced tremendously [2]. The ECG signal is used in various applications, including heart rate measurement, biometric identification, movement recognition, and anomaly diagnosis [3]. In general, ECG signals can be obtained by placing electrodes on the scalp with conductive gels [4]. The ECG is made up of potential fluctuations that are represented as an algebraic sum of cardiac fiber action potentials that can be computed from the body’s skin surface [5]. The typical structure of normal ECG signals aids in the detection of heart abnormalities, which are referred to as heart disorders [6]. Abnormalities in the ECG signal indicate the presence of an illness. It may be feasible to diagnose the diseases by comparing tested signals with healthy control signals [7]. The importance of this research lies in developing the performance of a device by reducing the delay in processing by minimizing the input vector ’s dimension and, as a result, the classifier’s