Abstract
In this paper, we present an alternative approach to perform spike sorting of complex brain signals based on spiking neural networks (SNN). The proposed architecture is suitable for hardware implementation by using resistive random access memory (RRAM) technology for the implementation of synapses whose low latency (<1μs) enables real-time spike sorting. This offers promising advantages to conventional spike sorting techniques for brain-computer interfaces (BCI) and neural prosthesis applications. Moreover, the ultra-low power consumption of the RRAM synapses of the spiking neural network (nW range) may enable the design of autonomous implantable devices for rehabilitation purposes. We demonstrate an original methodology to use Oxide based RRAM (OxRAM) as easy to program and low energy (<75 pJ) synapses. Synaptic weights are modulated through the application of an online learning strategy inspired by biological Spike Timing Dependent Plasticity. Real spiking data have been recorded both intra- and extracellularly from an in-vitro preparation of the Crayfish sensory-motor system and used for validation of the proposed OxRAM based SNN. This artificial SNN is able to identify, learn, recognize and distinguish between different spike shapes in the input signal with a recognition rate about 90% without any supervision.
Highlights
Probing motor cortical activity has recently received increased attention for the exploitation of human brain signals within Brain-Computer Interfaces (BCI)
We demonstrated the high potential of possible hardware embedded Spiking Neural Networks (SNN) for spike sorting of brain activity signals, relevant for the analysis of large-scale brain signals
We showed that these systems allow for fast adaptation to new input data and completely unsupervised operation, independently from the number of spikes in the input signal
Summary
Probing motor cortical activity has recently received increased attention for the exploitation of human brain signals within Brain-Computer Interfaces (BCI). Advanced microelectrode array (MEA) technologies (Spira and Hai, 2013) are unique and increasingly powerful tools to explore the central nervous system in detail Nowadays, they consist of hundreds or thousands of microelectrodes that allow recording the activity of large neural ensembles and especially spikes (action potentials) generated by the surrounding single cells (see Figure 1B). The offline processing is not optimal because it does not allow for real-time processing in closed-loop applications [e.g., in BCI (Hochberg et al, 2006, 2012)] or real-time data compression prior to wireless transmission with reasonable power consumption in case of high channel counts It was shown in Wessberg et al (2000) and Ifft et al (2013) that BCI performances are enhanced when recording from large numbers of neurons by means of large MEA’s, i.e., numerous signals have to be stored and decoded resulting in exploding data rates and computational efforts, respectively.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
