The Local Field Potential Signal Denoising And Feature Extraction Of The Mouse

As for brain information processing mechanism and brain-computer interface(BCI)research,it is significant to extract the brain wave signals that are associated with physiological information or behavior In order to obtain higher quality and more direct brain wave signals,implantable BCI gradually developed.Due to the fact that the local field potential(LFP)signal has a more stable long-term implantation,it has attracted the attention of many researchers in the research of implantable BCI.However,the LFP signal is recorded by a micro-electrode array,and it has small amplitude and a non-stationary signal,moreover,it is easy to introduce noise interference signals during the collecting process.All of these factors bring the difficulty to signal feature extraction and analysis of LFP in the latter stage.Therefore,good noise reduction process plays an important role before LFP feature extraction.In this thesis,the LFP signals are processed and analyzed through the Empirical Mode Decomposition(EMD)method based on the wavelet packet denoising.Firstly,the LFP signals are decomposed and reconstructed by wavelet packet.The noise signals were removed through this way.Then,the denoised LFP signals were decomposed through EMD method,and finally the characteristic wave is extracted.The main work content is as follows:1.The research focused on two kinds of common scalp EEG signal denoising methods-wavelet denoising and wavelet packet denoising,and the two denoising methods are applied to process LFP signal.Using MATLAB software to simulate,the power spectrum before and after signal denoising was obtained.The denoising evaluation index was established based on SNR,RMS error and related coefficient,and the denoising effect of the two methods was quantitatively analyzed.2.Wavelet denoising and wavelet packet denoising were combined with the EMD decomposition.Record the algorithm running time,the intrinsic modal component and the correlation coefficient of the LFP signal when performing EMD decomposition under the two denoising methods.The result indicates that both methods enable to reduce the boundary effect errors,decomposition layers of EMD decomposition and to improve its accuracy and timeliness.However,the comparison shows that the EMD decomposition and lifting effect based on wavelet packet is more obvious.3.Characteristic wave were extracted from the denoised LFP signals.The ?component were adopted from the LFP signals through the wavelet packet-based EMD decomposition method,and compared with the 8 component which is extracted from the wavelet packet transform method.The experiment proved that the former extracted the ? component more accurately.Finally,the correlation were analysed between the ? wave extracted through the cross-spectral density method and respiratory rate.The results showed that the ? wave extracted from mice LFP signal has a large correlation with respiratory frequency,which proves the accuracy and effectiveness of the algorithm used in feature extraction.