Grasp and hand-state decoding

Grasp decoding estimates the configuration and dynamics of the hand — discrete grasp types (power grasp, pinch, key grip), a continuous aperture or grip force, or individuated finger movements — as opposed to the endpoint trajectory of the arm. It is essential for useful manipulation: reaching to an object is worthless without the ability to close, hold, and release. Grasp can be decoded as a discrete state (classify among a small set of postures and trigger a pre-planned closing motion) or as a continuous signal (decode aperture or force directly), with the discrete approach more robust and the continuous approach more expressive.

Individuated finger control is substantially harder than whole-hand grasp because finger representations in motor cortex are overlapping and correlated, and natural hand movements are strongly synergistic. Progress has come from decoding into a low-dimensional synergy or postural-primitive space rather than commanding each finger independently, and from framing grasp as its own state dimension co-decoded alongside reach velocity. Dexterous, human-like hand control from cortex remains one of the harder open problems in motor BCI.