Comparison of decision tree algorithms for EMG signal classification using DWT. (April 2015)

• Record Type: Journal Article
• Authors: Gokgoz, Ercan; Subasi, Abdulhamit
• Is Part Of: Biomedical signal processing and control. Volume 18(2015)
• Page Start: 138
• Page End: 144
• Publication Date: 2015-04
• Subjects: Electromyography (EMG) -- Motor unit action potentials (MUAPs) -- Random forest -- C4.5 -- CART -- Multi-scale principle component analysis (MSPCA) -- Discrete wavelet transform (DWT); Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2014.12.005
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: Decision tree algorithms are used for EMG signal classification. EMG signals are de-noised using MSPCA, and decomposed into the frequency sub-bands using DWT. Combination of DWT and random forest achieved best performance with 96.67% total classification accuracy. Abstract: Decision tree algorithms are extensively used in machine learning field to classify biomedical signals. De-noising and feature extraction methods are also utilized to get higher classification accuracy. The goal of this study is to find an effective machine learning method for classifying ElectroMyoGram (EMG) signals by applying de-noising, feature extraction and classifier. This study presents a framework for classification of EMG signals using multiscale principal component analysis (MSPCA) for de-noising, discrete wavelet transform (DWT) for feature extraction and decision tree algorithms for classification. The presented framework automatically classifies the EMG signals as myopathic, ALS or normal, using CART, C4.5 and random forest decision tree algorithms. Results are compared by using numerous performance measures such as sensitivity, specificity, accuracy, F -measure and area under ROC curve (AUC). Combination of DWT and random forest achieved the best performance using k -fold cross-validation with 96.67% total classification accuracy. These results demonstrate that the proposed approach has the capability for the classification of EMG signals with a good accuracy. In addition, the proposed framework can be used to support clinicians for diagnosis of neuromuscular disorders.