M8 Brain-machine interfaces

Title

A brain-machine interface decodes motor or cognitive intention from recorded neural activity, directly extending the population coding and decoding formalism of E13. An adaptive linear decoder adjusts its weights using a supervised learning rule that minimizes the error between the decoded signal and the target intention, and this decoded signal controls an external effector in a closed-loop control scheme with artificial sensory feedback. Repeated use of the interface also induces cortical reorganization (extension of the plasticity of E07, along the same lines as M05), since the cortex learns to modulate its activity to improve control of the device. Interface performance is quantified by the information transfer rate (Wolpaw formula, based directly on information theory introduced in E13) and by decoding fidelity, which typically degrades with user fatigue over prolonged sessions.

ID:4280