This document presents a study on classifying ECG signals as normal or arrhythmic using wavelet-based feature extraction. The study used 35 1-minute ECG recordings from the MIT-BIH database, with 15 normal and 20 arrhythmic. Features extracted included Hjorth descriptors, entropy, and a proposed multi-feature set. Classifiers like ensemble KNN, linear SVM, and weighted KNN were applied, with the ensemble KNN achieving the best accuracy of 82.9% on the proposed feature set. The study concluded the feature set showed good classification performance but future work could explore other ECG signal attributes for more accurate arrhythmia detection.
Call for Papers - Educational Administration: Theory and Practice, E-ISSN: 21...
Â
Recognition of Arrhythmic Electrocardiogram using Wavelet Based Feature Extraction
1. Recognition of arrhythmic Electrocardiogram using
Wavelet based Feature Extraction
Authors
Atrija Singh
Dept. Of Electronics and Communication Engineering
Academy Of Technology
Debanshu Bhowmick
Department Of Applied Electronics and Instrumentation Engineering
Academy Of Technology
Subhadeep Biswas
Department Of Applied Electronics and Instrumentation Engineering
Academy Of Technology
3. Objective of the Study
⢠To develop a unique feature extraction approach to classify
a set of ECG signals into normal and arrhythmic set
r logo here
4. Dataset
⢠Collected from MIT-BIH ARRHYTMIA DATABASE
⢠Sampled at 360 Hz
⢠Considered 35 ECG recording of 1 minute duration
⢠The 15 recordings correspond to healthy Subjects while the rest are associated with
diseased(arrhythmic )Subjects.
⢠The signals were High-pass filtered using a 6th order Butterworth filter at cut off
frequency 0.5 Hz to remove the base line drift.
⢠Savitzky Golay filter were used for smoothening the ECG signal and removing any
noise.
logo here
ECG Signal
Acquisition
High Pass
Butterworth Filter
Savitzky Golay
Filter
Analog to Digital
Conversion at 360
Hz Sampling
Frequency
6. Previously Used Approaches on computer
based Arrhythmia detection
Daqrouq et al proposal
⢠Wavelet transform to recognize Arrhythmic ECG recordings
Rizel et al proposal
⢠Hjorth descriptor to classify ECG signal
Wachowiak et al proposal
⢠Analyzing multi resolution wavelet entropy with visual analytics
Balachandran et al proposal
⢠Daubechies algorithm for ECG signal feature extraction
logo here
7. Proposed Time Domain Multi-Feature Set
Proposed Multi Feature Set
â˘Hjorth Descriptor
â˘Entropy
logo here
9. Division of Dataset for Classification
⢠Training : 60%
⢠Validation : 40%
logo here
10. Results
logo here
Classification
Accuracy (%)
Feature Set Used
Set I Set IV
Ensemble(Subsp
ace K-NN)
81.8 82.9
Linear SVM 76.0 80.0
Weighted K-NN 74.3 77.0
Classification
Accuracy (%)
Feature Set Used
Set II Set IV
Ensemble(Subsp
ace K-NN)
63.6 82.9
Linear SVM 68.6 80.0
Weighted K-NN 66.7 77.0
Classification performance comparison with DWT
Coefficients(Set I) and Our Proposed feature Set IV
Classification performance comparison with HJORTH
Descriptor (Set II) and Our Proposed feature Set IV
11. logo here
Classification Accuracy
(%)
Feature Set Used
Set III Set IV
Ensemble(Subspace
K-NN)
79.9 82.9
Linear SVM 62.9 80.0
Weighted K-NN 74.3 77.0
Classification performance with entropy(Set III) and our proposed
feature Set IV
13. Conclusions and Future Scope
⢠Our feature set shows a good score of accuracy
with Ensemble(Subspace K-NN)Classifier
⢠Only R peak count cannot be considered as a
good scheme for disease detection.
⢠HRV can not be treated as the sole parameter to
classify arrhythmic ECG signals.
⢠We must calculate other attributes of the ECG
signals for better and accurate detection.
⢠This study can be further implemented for
classification and clustering of other bio-signals.
logo here
15. References
[1] K. Daqrouq and I. N. Abu-Isbeih, "Arrhythmia Detection using Wavelet
Transform," in EUROCON, 2007. The International Conference on
"Computer as a Tool", 2007.
[2] A. Rizal and S. Hadiyoso, "ECG signal classification using Hjorth
Descriptor," in Automation, Cognitive Science, Optics, Micro Electro-
Mechanical System, and Information Technology (ICACOMIT), 2015
International Conference on, 2015.
[3] M. P. Wachowiak, R. Wachowiak-Smolikova, D. J. DuVal and M. J.
Johnson, "Analyzing multiresolution wavelet entropy of ECG with visual
analytics techniques," in Electrical and Computer Engineering (CCECE),
2016 IEEE Canadian Conference on, 2016.
[4] A. Balachandran, M. Ganesan and E. P. Sumesh, "Daubechies algorithm
for highly accurate ECG feature extraction," in Green Computing
Communication and Electrical Engineering (ICGCCEE), 2014
International Conference on, 2014.
[5] G. Moody and R. Mark, " The impact of the MIT-BIH Arrhythmia
Database," IEEE Eng in Med and Biol, vol. 20, no. 3, pp. 45-50, 2001.
[6] S. P. M and S. E. M, "Analysis of ECG signal denoising using discrete
wavelet transform," in Engineering and Technology (ICETECH), 2016
IEEE international conference on ,2016
logo here