Leveraging Recurrent Neural Networks for Predicting Motor Movements from Primate Motor Cortex Neural Recordings
This work addresses brain-computer interface applications for motor decoding, representing an incremental improvement in efficiency and performance for a specific domain.
The paper tackled decoding motor movements from primate neural recordings using an Autoencoder Gated Recurrent Unit model, achieving a 0.71 R² score and ranking first in a benchmark challenge, with pruning reducing operations by 41.4% with minimal performance loss.
This paper presents an efficient deep learning solution for decoding motor movements from neural recordings in non-human primates. An Autoencoder Gated Recurrent Unit (AEGRU) model was adopted as the model architecture for this task. The autoencoder is only used during the training stage to achieve better generalization. Together with the preprocessing techniques, our model achieved 0.71 $R^2$ score, surpassing the baseline models in Neurobench and is ranked first for $R^2$ in the IEEE BioCAS 2024 Grand Challenge on Neural Decoding. Model pruning is also applied leading to a reduction of 41.4% of the multiply-accumulate (MAC) operations with little change in the $R^2$ score compared to the unpruned model.