The Advantage of Low-Delta Electroencephalogram Phase Feature for Reconstructing the Center-Out Reaching Hand Movements

摘要: It is an emerging frontier of research on the use neural signals for prosthesis control, in order to restore lost function amputees and patients after spinal cord injury. Compared invasive signal based brain-machine interface (BMI), a non-invasive alternative, i.e., electroencephalogram (EEG)-based BMI would be more widely accepted by above. Ideally, real-time continuous neuroprosthestic control required practical applications. However, conventional EEG-based BMIs mainly deal with discrete brain activity classification. Until recently, literature has reported several attempts achieving reconstructing movement parameters (e.g., speed, position, etc.) from EEG recordings, low-frequency band consistently encode motor information. Previous studies executed tasks have extensively relied amplitude representation such slow oscillations building models decode kinematic parameters. Inspired recent successes instantaneous phase sensory processing domains, this study examines extension slow-oscillation two-dimensional hand movements, first time. The data analysis are collected five healthy subjects performing 2D center-out reaching along four directions two sessions. On representative channels over cortices encoding execution information we show that low-delta characterized higher signal-to-noise ratio stronger modulation tasks, compared representation. Furthermore, tested commonly used linear decoding models. results demonstrate decoders lead superior performance reconstruction its counterparts, as well other-frequency power features. Thus, our contributes improve introducing new feature set patterns, demonstrates potential fine motion neuroprostheses.