A Novel Machine Learning Model for Assisting People with Motor Deficiencies

Abstract

Motor defects are a major problem affecting millions of people around the world. These individuals suffer from weakness in day-to-day functioning, which can lead to decreased and incoherent daily routines and impair their quality of life. This research describes a new machine learning-based model intended to help individuals with limb motor disabilities using their brain signals to control assistive devices in their daily life activities. The proposed model uses Empirical Mode Decomposition for removing the artifacts of the electroencephalography (EEG) signal, a modified Principal Component Analysis to reduce the input channels, and wavelet transform to extract features. In this experiment, discrete wavelet transform was used to decompose the signal at four levels. The approximate coefficient Ca and all level detail coefficients Cd4, Cd3, Cd2, and Cd1 were used to get the feature vector. All previous coefficients were used as input to Independent Component Analysis for feature reduction. Many amplitude estimators for neurological activities were defined mathematically to get the feature vector; finally, we classified the data using an artificial neural network. The proposed model evaluation was confirmed by testing on three different benchmark datasets, and the resulted accuracy of the proposed model was 88.067%, which outperforms a wide range of many current approaches.