An ECG Signal Classification System Using EMD and Pipelined Yolo Network

The objective of this work is to propose an efficient method using a deep learning approach for detecting abnormal electrocardiogram (ECG) signals. This article introduces a methodology for detecting arrhythmia ECG signals amidst normal ECG signals, targeting heart disease diagnosis. The proposed method applied Chebyshev filter on the source ECG signal to detect and remove the noise as a preprocessing method. The preprocessed ECG signals are decomposed into number of sub bands using Empirical Mode Decomposition (EMD) transformation method. The decomposed four sub bands are integrated into a Unique Matrix Pattern (UMP) which are further trained and classified using PYOLO architecture. The proposed PYOLO architecture classifies the source ECG signal into either normal or Arrhythmia. The proposed methods are tested on the ECG signals which are available in MIT-BIH Arrhythmia and SHEDB dataset. The method’s efficacy is validated using the MIT-BIH Arrhythmia and SHEDB datasets. The proposed EMD-PYOLO method for ECG classification, for SHEDB datasets, 97.5% Sensitivity, 97.5% Specificity and 97.5% accuracy and for MIT-BIH dataset ECG signals it obtains 99.8% Sensitivity, 99.7% Specificity and 99.8% accuracy. From the experimental result analysis, the proposed method is suitable for detecting the abnormal ECG signals from the normal ECG signals. Robustness could be improved in the future by using a variety of datasets and sophisticated preprocessing methods. Real-time ECG monitoring system integration could confirm the applicability in practice. Clinical optimization requires collaboration with medical specialists.