Brain-computer interface for amyotrophic lateral sclerosis patients using deep learning network

Individuals with Motor Neuron Disease were unable to move from one place to another because it gradually reduced all the voluntarily movement due to the degeneration of upper and lower motors neurons. The solution to this problem was to develop rehabilitating devices using biosignals. In this study, we have designed and developed electrooculogram-based wheelchair control using Cross Power Spectral Density. The convolution neural network to verify the performance and recognition accuracy of the wheelchair navigation in the indoor environment by using four trained users and four untrained users between the different age-groups and obtained the accuracy of 91.18% and 86.88% by using four fundamental tasks. From the indoor performance, the subject S4 from trained users outperforms all the trained subjects with an average classification accuracy of 93.51%. To verify the recognition accuracy, we conducted the online performance from the online performances subject S4 from trained subjects outperforms remaining trained subjects at the same time the subject S6 from untrained subjects outperforms all the untrained subjects. From the entire study, we analyzed that classification accuracy of subjects S4 was appreciated compared to other subjects. Through the research, we confirmed that the entire trained subject's performance was maximum compared to the untrained subjects in all the circumstances.