时间－频率联合分析方法可以揭示信号频谱的局域特征，是分析非平稳随机生理信号的有效手段。本文以脑电信号为例，介绍了常用于脑电数据处理的短时傅里叶变换、复小波变换和多正交窗3种线性时频分析方法及其优缺点，并进一步阐明时间－频率不确定性原理在时频变换中的关键作用。在具体实践中，将不同时频分析方法应用于经预处理后的任务态立体定向脑电信号，得到的时频谱均能有效揭示信号各频率成分的动态特征。进一步的对比表明具有可变Q值的Morlet复小波变换可以较好地提取信号的细节特征，是一种精确度较高且实现灵活的分析方法。最后，利用时频分析得到的能量谱计算用于反映脑间神经相似性的相关性指标，通过建立“时间－空间－频率”图像，说明时频变换在揭示大脑认知神经机制中的重要作用。本文是理解和完善电生理信号数据处理框架下时频分析方法体系的一次有效尝试。

Time-frequency analysis can reveal the temporal characteristics of the frequency spectrum of a signal, and is an efficient tool to analyze non-stationary and random electrophysiology signals. With respect to electroencephalogram, we introduced three commonly-used linear time-frequency analysis methods—short-time Fourier transform, complex wavelet transform and multi-tapers. We further illustrated the crucial role of time-frequency uncertainty principle in signal processing. For practical use, different time-frequency analysis methods were applied to the preprocessed task-related stereo-electroencephalogram signals. We confirmed the effectiveness of different time-frequency analysis methods in extracting dynamic characteristics of the frequency components. After comparing the results, more detailed information was provided by Morlet wavelet transform with a tunable Q value, which is a flexible method with higher precision. Finally, the power spectrum obtained from the time-frequency analysis was used to calculate the power correlation index to reveal inter-brain neural similarity. We established a “time-space-frequency” framework to interpret the significance of time-frequency analysis in unveiling the neural mechanism of certain cognitive process of human brain. Our work provided an attempt to understand and improve the framework of time-frequency analysis in electrophysiological signal processing.