Multi-channel neuro signal classification using Adam-based coyote optimization enabled deep belief network. (August 2022)

• Record Type: Journal Article
• Authors: Karunakar Reddy, Vanga; Kumar AV, Ravi
• Is Part Of: Biomedical signal processing and control. Volume 77(2022)
• Publication Date: 2022-08
• Subjects: Multi-channel neuro signal classification -- Electroencephalogram -- Enhanced deep belief network -- Adam-based Coyote optimization algorithm -- Deep learning; 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.2022.103774
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: This research work is highly focused on a multi-channel neuro signal classification model using developed deep architecture along with the help of suggested optimization algorithm for accurately detecting the brain diseases like epilepsy, brain tumor and Parkinson's disease. To develop an enhanced deep architecture named Enhanced Deep Belief Network (E-DBN) by tuning certain parameters of Restricted Boltzmann Machines (RBMs) layers in DBN using suggested Adam-based Coyote Optimization Algorithm (Adam-COA) for achieving high accuracy and precision while classifying the brain diseases. To introduce a new optimization algorithm named Adam-COA for optimizing the counts of the hidden neurons and learning rate of RBM layers in DBN to enrich the performance of classification of the proposed model. To evaluate an efficiency of the suggested multi-channel neuro signal classification model by comparing with existing meta -heuristic algorithms and classifiers and also performed overall evaluation on the developed method. Abstract: Electroencephalogram (EEG) signals gather the spiking activities of the brain based on its standardized electrodes of the scalp. Classification of EEG signal is a significant task for designing the precise "Brain-Computer Interface" system. Even though numerous studies have included the "time and frequency domain features" for classifying the EEG signals, a few studies combine the "spatial and temporal dimensions" of the EEG signal. Brain dynamics consists of high complexity among various mental tasks and so, it is challenging to build an efficient approach with features using their prior knowledge. Therefore, the main intention of this paper is to deal with the automated multi-channel EEG signal classification concerning different diseases like epilepsy, brain tumor, and Parkinson's disease. The EEG signals are initially gathered from the different benchmark datasets. Further, the feature extraction of signals is performed by the Haar DWT and spike detector. These features are further subjected to the "Enhanced Deep Belief Network" with RBM layers, in which the parameter tuning of DBN is performed by the Adam-based Coyote Optimization Algorithm for classifying the signal. The experimental results demonstrate the effective performance of the developed neuro signal classification model by reducing the signal complexity while classifying the signals at high accuracy.