Multi-Modal Integration for Motor Imagery EEG Signal Classification Combining with DNN and Functional Near-infrared Spectroscopy

Background: Motor perception Brain computer interface (BCI) applications have attracted significant attention to the classification of EEG signals. In contrast, the investigation of multimodal integration with additional neuro imaging modalities has been spurred by the drawbacks of employing the EEG signal alone, such as its low spatial resolution and noise susceptibility. In order to improve motor imaginary EEG signal classification performance.Methods: a novel method combining deep learning algorithms and functional near-infrared spectroscopy (fNIRS) is proposed in this research paper. The EEG signal used in this study is non-linear, complex, and non-stationary. Common spatial pattern (CSP) spatial filtering is used for preprocessing after the EEG is acquired. The quantity of data points must be divided up using wavelet extraction. Feature optimization with feature selection and simplified pulse coupled neural network (SPCNN) Classification used to more effective categorization of EEG signals.Results: The proposed method's results show an extraordinary accuracy of 96.87% when compared to other methods.