Ultra-Low-Power VLSI-Enabled Embedded Systems for Neural Prosthetics: Toward Scalable and Real-Time Brain–Machine Interfaces

Abstract

BMIs and neural prosthetics bring a world-revolutionizing promise in terms of defeating sensory and motor impairments in patients with neurological deficiencies. The current systems have severe drawbacks usually in the aspects of energy efficiency, effective interfacing with hardwares, and real-time capability, which renders them inappropriate in several wearable or implantable applications over a long period. The given paper suggests an ultra-low-power embedded architecture based on VLSI that has been especially tailored to overcome such limitations and facilitates elastic resource-constrained neural signal processing in real-time. The main aim of this study is to develop the compact, energy-efficient system-on-chip (SoC) platform which will consist of neural signal acquisition, spikes-based preprocessing, and embedded classification consumption of minimum power. The system is equipped with subthreshold analog front-end, hardware-efficient spike detectors, and a lightweight, embedded neural decoder combined with a dynamic power management unit, to reduce the energy requirements according to the demand of the workload. Simulation and FPGA-based prototype show that the system is capable of processing up to 64 channel of neural traffic at a rate exceeding 300 000 channels/second with average power-consumption below 120 micro-watts/ channel and a total processing-latency of 2.8 milliseconds. A comparative study shows major reduction in the amount of power consumption (~70%) compared to existing designs, without killing the signal fidelity or the decoder accuracy. This shows the functionality of the offered architecture and makes it appropriate to be implemented in next-gen neural prosthetic systems.In summary, the candidate VLSI-based embedded system offers a good way of creating ultra-low-power, real-time brain machine interfacing that can perhaps be used in an implantable biomedical device, adaptive neuro prosthesis, and closed loop neural control machine.