Deep neural network-based (BMI) brain-computer interface system design : Development and evaluation of a BMI brain-computer interface learning framework for rehabilitation

Based on the brain-computer interface is an important way of existing and future medical rehabilitation medicine, based on the very important role played by motor rehabilitation after stroke, the research framework based on reinforcement learning (RL) deep neural networks (BMI) brain-computer interface decoder utilizes an internal reward decoding framework and metrics to transform the mapping relationship between motor and rehabilitation. The brain is allowed to travel a customized set of learning mechanisms to identify the rewarding process with the correct behavior in a new task. This internal guidance can be used as an external reference. Data systems research can improve rehabilitation efficiency by 9%. The accuracy rate is up to 92%. And as a separate research and treatment modality. This study is based on the design of a deep neural network-based (BMI) brain-machine interface system, which is experimented with in both online and offline research directions, with 9 subjects (5 male, 4 female), aged between 25 and 34 years old, wearing 32-guided joint electrode caps on their heads, all with a resistance of 5kΩ or less. The glasses were 40cm~90cm away from the screen, the chief operating range and procedure were set in advance, and the sampling frequency of the Neuroscan-NuAmps EEG acquisition instrument was set at 250Hz. the tests were verified and improved. The research results of this paper have been applied for a patent and granted.