Action Intention Recognition and Robotic Arm Actuation Based on Bidirectional Long Short-Term Memory and Brain-Computer Interface

To recognize action intention of the people with physical disabilities or paralysis, the combination of braincomputer interface (BCI) technology and robotic arms is a solution with great potential. As one of the most common BCI paradigms, Electroencephalogram (EEG) motor imagery (MI) can capture the electrical signals in the human brain. It can classify the people’s intention hidden in the EEG signal and drive the robotic arm to realize the action intention. Deep learning (DL) methods are used to accomplish this classification task. In this paper, a BiLSTMbased approach is proposed to classify EEG signal, and then the classification result is imported into a robotic arm to catch or release an object.