Research On Spike Signal Detection And Clustering Algorithm

Neighboring neurons communicate with each other by launching and receiving spikes.Therefore,recording and analyzing neuronal spike train is an important method for neuroscience research.The extracellular recording electrode used in the early days can only carry one recording site,and the raw data obtained is relatively simple to process.With the development of electrode manufacturing technology,the widely used high-density electrode array can carry thousands of recording sites,and the distance between adjacent sites is extremely small.One spike can be recorded by multiple sites at the same time.The scale of data generated by the high-density electrode array and the processing difficulty have increased sharply.This paper chooses the raw data generated by the high-density electrode array as the research object,focusing on the spike extraction and clustering process of spike sorting.The main work done includes:1.This paper proposes a high-density electrode array extracellular recording model.In this paper,the relative position of neurons and the electrode array is determined according to the amplitude fluctuation of the spikes on different channels.The relative distance between the recording sites is calculated according to the electrode geometry,and then the attenuation direction of the spike amplitude is determined.In a word,the way how high-density electrode works is described in detail.2.Aiming at the spike detection and extraction,this paper proposes a local spike extraction algorithm.This paper introduces two parameters,main channel and spike span,and with the help of them we clarify the location information of the spike,adds a location feature to the extracted spike,solves the problem of homogenization of the results of overlapping spikes,and eliminates side effects of spikes on other channels,which improves extraction efficiency on other channels.3.Aiming at the spike clustering process,this paper proposes a channel-by-channel clustering algorithm.In this paper,according to the spatial feature of spikes,the extracted spikes are grouped,and the spikes are clustered in the same main channel.At the same time,to eliminate the side effect of overlapping spikes on subspace choice,we mark the spikes as overlapping spikes and non-overlapping spikes,and cluster them seperately.Based on the above three points,this paper proposes a systematic and complete spike sorting algorithm,and the feasibility and effectiveness of this algorithm are verified through experiments.