The Application Of Time-varying Autoregressive Model In Feature Analysis Of Local Field Potential In Awake Rats

Primary visual cortex (V1region) is a very important visual area, so depth analysis of the responsive characteristic for the signal in V1is highly significant for people to understand the visual pathway of rodents and the role of the visual pathways in the visual cognition. Synchronized and desynchronized are two typical states in sober rats’ brain activity, which are also the basic of analyzing differences of time-frequency for Local Field Potentials (LFP) under different states and brain information processing mechanisms. Time-frequency spectrum analysis could be regarded as efficient methods when time and frequency domain information of Local Field Potentials are studying. Traditional method of frequency analysis can’t provide a high precision for its low time-frequency resolution or computationally intensive problems and can’t distinguish the energy change of Local Field Potentials accurately under states of synchronous and desynchronized. Therefore, Time-frequency spectrum analysis based on time-varying autoregressive model combined Kalman filtering algorithm attracts many researchers gradually.Based on the analysis, this thesis, which is aiming at analysis the nonlinear characteristics of Local Field Potentials for conscious rats under states of synchronous and desynchronized, completes the comparisons of the analysis of real-time spectrum which are based on time-varying autoregressive model and Kalman Filter Algorithm, Fourier Transform and Wavelet Transform, and verify its superiority. The main contents are as follows:1. Characteristics of chronic experiments and signal. Local Field Potentials Signals, whose characteristics are analyzed under states of synchronized and desynchronized, are obtained through chronic implant surgery and recording sober LE rats in V1. Finally, the characteristic of Local Field Potentials Signals are studying according to commonly analysis methods, which is the base of priority knowledge for researching time-frequency characteristics of Local Field Potentials Signals.2. Research of traditional time-frequency spectrum analysis. Time-frequency spectrum based on the method of short-time Fourier transform (STFT) is studying firstly, and time-frequency spectrum analysis using STFT method can’t give higher precision frequency resolution with the restrict of "Uncertainty Principle". Time-frequency analysis based on Wavelet Transform can increase the time-frequency resolution, but the choices of wavelet function is difficult, and because of problem of the phase,the accuracy of time-frequency analysis is very low.3. The establishment of state space model for Local Field Potentials. TVAR model spectrum analysis method is used in order to improve the accuracy of dynamic spectrum analysis. Parameters are updated iteratively by Kalman filter. Compared with traditional TVAR model coefficient estimation, Kalman Filter Algorithm achieves the optimal model, and the means of estimation error and variances are small.4. Comparison of effect for time-frequency spectrum analysis methods. The research is based on comparison of the time-frequency spectrum analysis of TVAR model combined with Kalman Filter Algorithm and the time-frequency spectrum analysis of STFT and wavelet transform, and analyzes resolutions of the three methods, significant differences and correlation coefficient for Local Field Potentials. The result shows that time-frequency spectrum analysis based on TVAR model which is combined with Kalman Filter Algorithm has higher time-frequency resolution, and can distinguish the significant differences of Local Field Potentials in the states of synchronized and desynchronized, and has higher correlation coefficient. Therefore, the approach based on TVAR model has more advantages for Local Field Potential in the Real-Time Spectrum Analysis.