Early movement intention recognition via corticomuscular synchrony in complex tasks

Intuitive control of assistive robots requires precise decoding of movement intention before muscle activation. While corticomuscular coherence (CMC) tracks descending corticospinal drive, decoding pre-onset intention during complex, multi-target tasks remains challenging. This study identifies stable CMC markers characterizing the transition from motor planning to execution.

Twelve healthy volunteers performed target-directed reaching tasks using a 6-DoF robotic arm controlled via a wireless wearable inertial interface. The experimental protocol evaluated two distinct conditions: a four-target reaching task arranged in a semi-circle and a more complex two-target reaching task with 3D obstacles requiring trajectory planning. An exploratory analysis was performed using wavelet transform coherence (WCT) to map the pre-movement time-frequency dynamics and simulated real-time framework was subsequently implemented to classify 'rest' and 'intention' states, comparing CMC metrics against EEG power spectral density (PSD).

Exploratory time-frequency analysis using WCT revealed a statistically significant pre-motor desynchronization across the high theta, alpha, and beta bands, which was concentrated predominantly 100 to 200 ms prior to EMG activation. In the classification framework, the metric leveraging phase dynamics in CMC outperformed the PSD classifiers, achieving a mean accuracy of 0.80 accross participants and highest peak of intention classification around 100 ms before EMG onset.

Corticomuscular coherence provides a highly reliable and very early neurophysiological marker of higher-order motor planning and sensorimotor integration before physical execution. Integrating these phase-desynchronization features into future real-time multi-sensor fusion architectures offers a viable strategy to enhance intent recognition accuracy and shorten system latencies in assistive rehabilitation.