Madhavan P G, Stephens B E, Klingberg D, Morzorati S
Department of Electrical Engineering, Purdue University, Indianapolis, IN 46202.
J Neurosci Methods. 1991 Dec;40(2-3):91-100. doi: 10.1016/0165-0270(91)90057-7.
The main objective of this paper is to make the auto regressive (AR) power spectrum estimation method accessible to electrophysiologists and present some typical applications. The AR method is explained, choices of various parameters are explored and examples from the analysis of rat hippocampal EEG are given. We also provide the pseudocode for the computation of the AR coefficients and the AR spectrum. We compare the results from the AR method to the FFT-based power spectrum method and demonstrate the superiority of the AR method. We also show the differences in the spectra of the EEG of alcohol-preferring (P) and non-preferring (NP) rats in the baseline condition when no alcohol is infused. We found a statistically significant difference in peak theta frequency which was at 6.96 Hz for the P rats and at 7.74 Hz for the NP rats. There were also other observable differences in the shape of the spectra of the EEG of the P and NP rats.
本文的主要目的是让电生理学家能够使用自回归（AR）功率谱估计方法，并展示一些典型应用。文中解释了AR方法，探讨了各种参数的选择，并给出了大鼠海马脑电分析的示例。我们还提供了计算AR系数和AR谱的伪代码。我们将AR方法的结果与基于快速傅里叶变换（FFT）的功率谱方法进行比较，证明了AR方法的优越性。我们还展示了在不注入酒精的基线条件下，嗜酒（P）大鼠和非嗜酒（NP）大鼠脑电谱的差异。我们发现，P大鼠的θ波峰值频率在6.96Hz，NP大鼠的θ波峰值频率在7.74Hz，两者存在统计学上的显著差异。P大鼠和NP大鼠脑电谱的形状也存在其他可观察到的差异。
