Search Results for Abdulhamed M. Jasim

Abstract

Since cardiac conditions are among the most fatal illnesses in the medical community, ECG classification systems are crucial for understanding and diagnosing patients’ health conditions. Numerous techniques for ECG feature extraction and classification algorithms are developed by researchers. This paper presents a method for accurately classifying ECG illnesses based on the 3-scale Slantlet transform (SLT) and artificial neural network (ANN). The ability of the SLT filters to decompose the ECG signal at various resolutions led to excellent classification. As a new realization, all coefficients of the modified designed SLT filters are expressed by the sum-of-power-of-two (SOPOT) approach to reduce the complexity. It is noteworthy that the average and maximum deviation error values between the responses of original and modified filters are very small. Hardwarely, the new realization leads to a less complex implementation for the designed SLT filters on FPGA kit using the Xilinx System Generator for DSP with very small errors between output resposes of the original and modified filters. FPGA results show that the system is designed using a best-selected wordlength method. The proposed classification system is capable of distinguishing the ECG normal case and other four different diseases with a high overall accuracy of 98.50 %.