Decoding upper limb kinematics from primary motor cortical representations for intracortical brain-machine interfaces

We will present our recent study to investigate decoding of kinematic information from primary motor cortical firing activities for the control of upper limbs during arm reaching tasks in non-human primates. Decoding models include the construction of a low-dimensional manifold space representing population activity. Our study employs factor analysis and its variants to create such representational spaces. Conventional types of decoders such as linear filters predict the hand velocity from neural representations. Decoding performance results show that neural representations yield relatively better decoding than original firing rates for dissimilar reaching tasks. Our investigation may help to design more appropriate decoding methods for intracortical brain-machine interfaces for controlling upper limb prosthetics.