Brain-Computer Interface Based Control for Disabled

Brain-computer interface (BCI) has gained popularity in recent years as an alternate communication mechanism between the human brain and technology for output. WHO accounts for 15% of the world’s population with disabilities, and there is a critical need for assistive systems to aid those persons in carrying out their everyday activities. This study can help disabled people to increase their independence and quality of life. Electrical impulses from the human scalp are detected and converted into control commands that may be utilized to control external equipment, computer screens, and virtual objects in real time. The present work uses the Deep Learning method to interpret motor imagery signals for Brain-Computer Interface applications. ERD patterns are also gathered that represent the activation and deactivation of the motor cortex. In addition, task-related band power changes are recorded in every trial, which can show stimulation of sensory-motor areas. Furthermore, the goal is to categorize EEG data derived from a Motor Imagery task into two classes, viz. left and right, using the convolutional neural networks. The time-frequency pictures of EEG data fed as input to the convolutional neural network were obtained using continuous wavelet transform. The three channels were separated using data segmentation to convert the raw data into images. To get over the issue of few trials in motor imagery tasks, transfer learning strategies are employed. Transfer learning is applied to three premade neural networks GoogLeNet, AlexNet, and ResNet50, to proceed with CNN classification. Upon analyzing results, AlexNet achieved the best accuracy among the others.