| Visual system is the main channel for animals to obtain outside information, it is also significance for the blind men to able to see.Obtaining neuronal signals related to visual stimuli is the first step of the study for the visual system,and spikes are the basic form of neural network activity. Thus, studying spikes play an important role in further understanding how the visual system code,express and process the outside what do we see. As spikes are broadband, high frequency and small value, and are sensitive to noise. Therefore, noise interference suppression is the basis for obtaining high signal noise ratio of spikes and for spikes’ further study.In this paper,we firstly make power spectrum analysis of the spikes and noise signals of V1, according to their properties, we successively use the spike simulation signals and the experimental signals for researching and verifying, the research of relatively independent spikes is the main focus on the signal correlated noise suppression, and make quality evaluation of the noise suppression results. The main contents are as follows:1、We briefly analyse the generation and features of spikes,and the mode of stimulation in the experiment,and the experimental signals. Also,we respectively make analysis of their properties in time domain and frequency domain. The results shows that the experimental spike signals contain not only white noise,but also correlated noise, what’s more,there are many small value spike signals.2、Following common noise sources, we use two kinds of de-noising algorithm, PCA denoising and PCA-combined wavelet denoising algorithm.And we respectively use the simulation signals and the experimental signals to study and check the two denoising algorithm, their results show that,as for correlated noises, PCA denoising algorithm is not significant, as signal-to-noise ratio improves so littler, but the PCA-combined wavelet denoising algorithm can effectively improve the signal-to-noise ratio,however,as it use PCA denoising in front of the wavelet threshold denoising, although this denoising algorithm can improve the inhibition ability of wavelet to correlation noises, the PCA denoising will make the waveform of spikes distortion increasely.3、Aimed at the existing problems of the above algorithms, this paper uses the simulation signals and the experimental signals to study and validat multivariate wavelet denoising algorithm. In this paper, in essence, multivariate wavelet denoising algorithm is PCA-combined wavelet denoising algorithm, what the difference is this combination is done in the wavelet threshold denoising.In fact,the singular value decomposition of the first layer of Wavelet detail coefficients is as the use of principal component analysis which make the first layer of Wavelet detail coefficients into the new space, then different detail coefficients are projected to the space as for threshold denoising.In additiori,the introduction of detail coefficients amplitude magnitude of the threshold is useful,as it not only can reduce the distortion of the spike waveform but also improve the effect of denoising.4、In this paper, as for the simulation signals and the experimental signals results of three denoising algorithm, PCA and PCA-combined wavelet denoising and multivariate wavelet denoising algorithm, we make quality evaluation for them, and make comparative analysis of the interference suppression effects of the three kinds algorithms. The results show that multivariate wavelet denoising algorithm which can keep the suppression effect of white noise is very efficient, as at the same time it also can improve the removal effect of correlated noise, and effectively improve signal-to-noise ratio of neurons spikes. Simulation and experimental data show that the improvement of signal-to-noise ratio of the simulation signals is at about4.25, the experimental data is at about2.75,and the decrease of the distortion of spikes waveform the simulation data is at about3.9, the experimental data is at around2.8. |
PDF Full Text Request