Long-term asynchronous decoding of 3D hand trajectories using electrocorticographic signals in primates Toward a chronic asynchronous brain-machine interface

Brain-machine interfaces (BMI) based on electrocorticography (ECoG) utilize higher fidelity signals over non-invasive BMIs, and provide greater long-term stability over other invasive BMIs. However, little study on ECoG-based BMIs has focused on the asynchronous decoding of continuous actions that do not require external cues for start or stop. Here, we proposed a novel wavelet-based algorithm, and successfully decoded 3D continuous hand trajectories in a primate with accuracy comparable to that found in BMI studies using single unit activity. Furthermore, the performance was found to last for at least 2 months. Evidence of high accuracy and long-term stability elucidates the feasibility of chronic asynchronous ECoG-based BMIs.