Classification of biological signals and time domain feature extraction using capsule optimized auto encoder‐electroencephalographic and electromyography

Abstract

Electroencephalographic (EEG) and electromyography (EMG) signal classification seem to be a modulus topic in engineering and the medical field. The nature of the EEG and EMG signal is non-stationary, noisy and high dimensional. The intrusion of noise in the signal may distress movement recognition. A novel methodology is being developed in this research to deal with these issues. Here, the EEG and EMG signals are recorded using the BCI2000 system. The proposed model comprises three phases: pre-processing, feature extraction (FE), and motion classification. The pre-processing method can be used to enhance visual appearances and the quality of the signal. The hybrid discrete wavelet based delayed error normalized least mean square error (DWT-DENLMS) is introduced to eliminate the presence of motion artifacts in the recorded EEG and EMG signal. After pre-processing, the recorded signals are combined and forwarded to the FE stage. The hybrid Dual tree complex wavelet transforms based on Walsh Hadamard transform (DTCWT-WHT) is proposed to extract the indispensable time domain features from the combined biological signal. The hybrid capsule transient autoencoder (HCTAE) algorithm is proposed to classify the motion recognition (T0-rest, T1-left fist and both fists, and T2-right fist, both feet). The error in the network model is diminished by transient search optimization (TSO) strategy.The Python platform is used to implement the developed approach, and the performance of the optimized classification approach yields accuracy, precision, recall, F1 score and specificity of 98.51%, 97.25%, 97.94%, 97.58% and 98.94%, respectively.

BiomacEMG: A Pareto-Optimized System for Assessing and Recognizing Hand Movement to Track Rehabilitation Progress

“…The approach requires only 100 impulses out of each gesture for training, significantly reducing the time necessary to train the system. Such techniques suggested combine hand-crafted characteristics from a time-spectral investigation with deep features to build the feature vector [35]. Subsequently, the EMG characteristic vector is often classified using a multilayer perceptron classifier (MLPC) [36].…”

Toward Improved Classification of Perceived Stress Using Time Domain Features

“…Two varieties of spatial convolution kernels are incorporated, with a specific focus on highlighting disparities between brain hemispheres for spatial feature extraction. Neeraj et al extracted time domain features from the combined biological signal to classify them [21]. Techniques for extracting statistical features in the time domain, including mean correlation, kurtosis, and skewness, are employed for classification using the KNN classifier [22].…”