Neural signal recording and processing systems play a crucial role in understanding brain function and developing Brain-Computer-Interfaces. Specifically, spike sorting, which cluster the spikes based on their shapes, are important neural signal processing procedure identifying the source neurons of spike events. Traditional neural recording systems use Nyquist ADC to get uniform samples in time domain. However, since spikes occur as sparse events, there is a potential for power saving by utilizing event-based ADCs such as level crossing ADCs. While level crossing ADCs offer a power advantage, they introduce challenges to digital processing systems due to the non-uniformity of samples, often necessitating high-speed offline timers for signal reconstruction. This research endeavors to harness the power advantage of level crossing ADCs while mitigating the burden on digital processing systems. We propose a novel approach that eliminates the need for offline timers by directly using level-crossing-based pulse signals for on-chip spike detection and feature extraction. Furthermore, we introduce new features and optimized clustering algorithms to enhance spike sorting performance. The seminar will commence with an introduction to spike sorting and level crossing ADCs, followed by an in-depth exploration of the level-crossing-based sorting algorithm. Finally, we will present a comprehensive overview of the overarching architecture of the recording and sorting system.
31/5/2024 11:00 - 12:00
ESAT Aula L
