Spiking Neural Networks Based on OxRAM Synapses for Real-Time Unsupervised Spike Sorting.

Area of Science:

• Neuroscience
• Materials Science
• Computer Engineering

Background:

• Spike sorting is crucial for analyzing complex neural signals.
• Conventional methods face limitations in real-time processing and power efficiency.
• Brain-computer interfaces (BCI) and neural prosthetics require advanced signal processing techniques.

Purpose of the Study:

• To present a novel spiking neural network (SNN) architecture for efficient spike sorting.
• To leverage resistive random-access memory (RRAM) technology for synapse implementation.
• To demonstrate the potential for real-time, low-power neural signal processing.

Main Methods:

• Developed an SNN architecture utilizing Oxide RRAM (OxRAM) for synapses.
• Employed an online learning strategy inspired by Spike Timing Dependent Plasticity.
• Validated the SNN using intra- and extracellularly recorded spiking data from a Crayfish sensory-motor system.

Main Results:

• Achieved low-latency (<1μs) and ultra-low power (nW range) spike sorting.
• Demonstrated RRAM synapses with low energy consumption (<75 pJ) and ease of programming.
• The OxRAM-based SNN achieved approximately 90% recognition rate for different spike shapes without supervision.

Conclusions:

• The proposed OxRAM-based SNN offers a promising alternative for real-time spike sorting.
• This technology enables potential advancements in BCI, neural prosthetics, and autonomous implantable devices.
• The system demonstrates effective unsupervised learning and recognition of neural signals.