Trajectory Decoding of Arm Reaching Movement Imageries for Brain-Controlled Robot Arm System

Area of Science:

• Neuroscience
• Robotics
• Biomedical Engineering

Background:

• Noninvasive brain-machine interfaces (BMIs) using electroencephalography (EEG) are crucial for controlling external devices and neuro-rehabilitation.
• Spontaneous movement intentions can drive these BMI systems for enhanced user control.

Purpose of the Study:

• To present a brain-controlled robot arm system utilizing electroencephalography (EEG) for arm trajectory decoding.
• To investigate the feasibility of decoding movement intentions for intuitive robot arm control.

Main Methods:

• Developed an experimental system to acquire EEG data during movement execution (ME) and movement imagery (MI) tasks.
• Proposed a subject-dependent deep neural network (DNN) architecture employing bi-directional long short-term memory (LSTM) for robust arm trajectory decoding.
• Conducted experiments with five subjects performing four directional reaching tasks in a 3D plane.

Main Results:

• Achieved high decoding performance (r-value > 0.8) for X, Y, and Z axes in both MI and ME tasks across all subjects.
• Demonstrated consistent decoding performance between ME and MI tasks in offline analysis.
• Confirmed the feasibility of EEG-based intuitive robot arm control for complex tasks like object manipulation.

Conclusions:

• The proposed subject-dependent DNN model enables robust EEG-based arm trajectory decoding.
• This technology shows significant potential for real-time, intuitive control of robot arms in assistive and rehabilitative applications.
• The system's capability for robust trajectory decoding is promising for real-world applications.