Restoring natural reach-and-grasp kinematics in tetraplegic patients requires continuous multidimensional motor decoding. Latency delays between intention and prosthetic articulation above 50 milliseconds break the illusion of agency. This paper presents low-latency Kalman filter decoders running on FPGA edge gates.
Recursive steady-state Kalman filters update 3D arm position and hand aperture estimates every 10 milliseconds, dynamically adjusting noise covariance matrices to compensate for electrode signal dropout.
Prosthetic fingertips equipped with piezo-resistive pressure sensors transmit tactile data back to cortical sensory arrays via intracortical microstimulation (ICMS), completing the closed biological loop.