Research And Implementation Of Interference Suppression Algorithm To Neuronal Spikes

Visual system is one of the most abundant channels to perceive the outside environment as well as the most complex sensory system in terms of structure and function. In visual system, spikes are the main carrier to take and transfer visual information. As a result, the accurate detection of the spike is the premise to study the visual information processing mechanism. However, the spike data is susceptible highly to noise because it’s non-stationary, high-frequency, and small amplitude signal. Therefore, the research on the algorithm to effectively reject noise from the raw spike data and to improve the signal-to-noise ratio (SNR) plays key role in the study of spike information encoding.In this paper, we focused on the algorithm of interference rejecting to the raw spike data and single-unit quality assessment after detection or clustering based on statistical features of the spike and the noise. Finally, the algorithm was verified by simulating and actual data collected from ratⅥ. The main results as follows:1:Generation mechanisms and features of the spike were analyzed firstly. The types of background noise and their sources in the raw spike data were discussed, based on which the anti-interference system was established include shielding and grounding. The raw data was recorded after designing the experiment in grating stimulation.2:The statistical features of the spike and the noise were researched, and the results show that it not only contains Gaussian white noise, but also color noise in background noises. Based on the results the multi-variant de-noising algorithm was applied to improve the SNR of spike data based on wavelet transform and principal component analysis (PCA). PCA was used to kill insignificant principal components to obtain an additional de-noising effect. The algorithm was verified through the simulation and actual data.3:The sources and features of the noise with large amplitude were analyzed with the actual data and confirmed with the simulation data. According to these features, the de-noising algorithm based on adaptive interference cancellation was firstly proposed to remove the larger amplitude color interference and validated through the actual data. And the reference input in adaptive interference cancellation was obtained by common average reference, the feasibility of which was discussed using Gauss-Markov theorem and the actual data.4:In most cases, the quality assessment is based on the subjective judgment of human observers or signal-to-noise ratio, and the recorded units are divided into "well isolated" or "multi-unit" groups. This subjective evaluation precludes comprehensive assessment of single-unit studies since the quality of single-unit is not explicitly defined. In the paper, a novel quality evaluation algorithm following the idea of neighbourhood component analysis based on correlation distance was proposed to improve the stability and accuracy of estimation results. It would provide a reliable evaluation index for the following research on spikes.