Self-Supervised Learning Framework for Neural Decoding in Brain-Machine Interfaces
A new research paper proposes Self-Supervised Consistency enhanced Disentangled Learning (SSCDL), a framework to address neural drift in Brain-Machine Interfaces (BMIs). Neural drift causes performance degradation over time, especially in invasive BMIs. Existing methods struggle with robust neural representations and fail to account for drift varying across motor parameters like velocity, direction, and speed. SSCDL introduces a Consistency enhanced Neural Decoder (CND) using a teacher-student consistency mechanism. The paper is published on arXiv under ID 2607.24023.
Key facts
- SSCDL is a neural decoding generalization framework for BMIs.
- Neural drift reduces BMI performance over time.
- Existing solutions have two main drawbacks: difficulty in learning robust neural representations and neglecting drift variation across motor parameters.
- CND uses a novel teacher-student consistency mechanism.
- The paper is on arXiv with ID 2607.24023.
- BMIs enable control of assistive and robotic technologies.
- Applications include rehabilitation, human motor augmentation, and human-centered robotics.
- The framework targets invasive BMIs (iBMIs).
Entities
Institutions
- arXiv