Spike Sorting Chip Applied To Brain Computer Interface

The extracellular recording technique in neuroscience,the technique of inserting electrodes into the extracellular tissue of the brain to record the activity of individual neurons,is a common means of study brain function.This technology has great potential value in the fields of brain-computer interface,epilepsy and memory loss.For neural signal processing,spike sorting is a key step.spike sorting is a process of separating signals into individual neuron activities.Many related algorithms have been proposed for many years,but the hardware implementation has not been complete so far.In this paper,we study the realization of spike sorting IP,the main work is as follows:In this paper,the characteristics of the spike signal and the spike sorting process are analyzed.Among them,the signal pre-emphasis,spike detection,alignment,feature extraction,dimensionality reduction and clustering algorithms in the peak potential classification process are analyzed by hardware-implemented algorithms.In this paper,the spike sorting IP is designed and implemented,and its function is simulated and verified.The calculation function in each process of spike sorting is realized.The pre-emphasis module has been tape-out under the TSMC 180 nm process.This paper implements an ARM-based SoC system that applies spike sorting IP.The chip system can realize two system startup modes and design an external largecapacity storage device controller,which respectively satisfies the different working modes of the spike sorting IP and the storage of the spike signal data.This paper has performed simulation verification and board level verification on IP and system.This includes simulation and board level verification using both real and random spike signal data.The experimental results meet the design requirements.The spike sorting IP and SoC chips designed in this paper have a good application prospects and can be widely used in applications where spike signal hardware acceleration classification and compression data are required.