Improved Spike-Based Brain-Machine Interface Using Bayesian Adaptive Kernel Smoother and Deep Learning

Multiunit activity (MUA) has been proposed to mitigate the robustness issue faced by single-unit (SUA)-based brain-machine interfaces (BMIs). Most MUA-based BMIs still employ a binning method for estimating firing rates and linear decoder decoding behavioural parameters. The limitations of lead suboptimal performance BMIs. To address this issue, we propose which consists Bayesian adaptive kernel smoother (BAKS) as rate estimation algorithm deep learning, particularly quasi-recurrent neural network (QRNN), algorithm. We evaluated reconstructing (offline) hand kinematics from intracortical data chronically recorded primary motor cortex two non-human primates. Extensive empirical results across recording sessions subjects showed that consistently outperforms other combinations Overall suggest effectiveness improving