ArmBCIsys: Robot Arm BCI System With Time-Frequency Network for Multiobject Grasping

Area of Science:

• * Neuroscience
• * Robotics
• * Signal Processing

Background:

• * Brain-computer interfaces (BCI) offer communication channels for individuals with physical disabilities.
• * Decoding electroencephalography (EEG) signals, especially under low signal-to-noise ratio (SNR), is challenging for complex tasks like multiobject grasping.
• * Existing systems lack robust decoding algorithms and precise visual tracking for real-world applications.

Purpose of the Study:

• * To develop an integrated robotic arm system (ArmBCIsys) for multiobject grasping using noisy EEG signals.
• * To enhance EEG signal decoding robustness under low SNR conditions.
• * To improve visual tracking and segmentation for reliable object grasping.

Main Methods:

• * Proposed a novel Dual-Branch Frequency-Enhanced Network (DBFENet) with Scaling Temporal Convolution Blocks (STCB) and DropScale Projected Transformer (DSPT) for EEG feature extraction.
• * Integrated a vision-guided grasping module (VisGraspSeg) using fine-tuned Mask2Former and multiframe centroid-intersection over union (IoU) tracking.
• * Validated the system on a self-built and two public code-modulated visual evoked potential (c-VEP) datasets.

Main Results:

• * DBFENet achieved state-of-the-art recognition performance on c-VEP datasets.
• * The integrated ArmBCIsys demonstrated stable multiobject selection and automatic grasping in dynamic environments.
• * The system effectively decodes low SNR EEG signals for precise robotic arm control.

Conclusions:

• * ArmBCIsys provides a robust solution for controlling robotic arms via BCI, even with noisy EEG data.
• * The developed DBFENet and VisGraspSeg modules significantly advance BCI capabilities for assistive robotics.
• * This technology holds significant promise for healthcare robotics, assistive devices, and industrial automation.