Search papers, labs, and topics across Lattice.
This study introduces an event-based gated recurrent unit (GRU) for neural decoding in neuroprosthetic motor control, addressing the challenges of latency, energy consumption, and spatial requirements in existing systems. By generating a sparse communication pattern with graded spikes, the proposed method significantly enhances task performance compared to classical spiking neural networks. The efficient training and sparse inference techniques employed enable practical on-device implementation, paving the way for improved mobility solutions for patients with disabilities.
Event-based neural decoding achieves superior task performance while drastically reducing energy consumption and latency, revolutionizing neuroprosthetic control.
A substantial number of patients experience diminished mobility due to disabilities, diseases, or accidents. Although modern prostheses, powered by deep neural networks, hold the promise of significantly enhancing the quality of life for these individuals, their widespread adoption is hindered by significant latency, energy consumption, and spatial requirements. Wired connections to external high-performance processors restrict patient mobility, while wireless connections limit the volume of information that can be transmitted to these processors. Spiking neural networks offer the potential for compressed communication and low-power inference, yet they often lag behind state-of-the-art deep learning models in various applications. In this study, we propose a high-performance neural decoding method that effectively balances task performance and efficiency. An eventbased gated recurrent unit generates a sparse communication pattern with graded spikes, surpassing classical spiking neural networks in terms of task performance. Utilising an efficient training method and sparse inference, our model presents new opportunities for on-device neural decoding.
