Spectral and time-frequency analysis

EEG signals are typically characterized by oscillatory patterns at certain frequency bands. Normally, the EEG data, especially spontaneous EEG data, are analyzed in the frequency domain. The spectral analysis can transform EEG signals from time domain to the frequency domain, which can reveal how the power of EEG signals is distributed along frequencies. Furthermore, as EEG spectrum could substantially vary over time, joint time-frequency analysis is often used to reveal time-varying spectral activities of EEG. Particularly, time-frequency analysis is a powerful method to estimate the event-related EEG spectral patterns, i.e., eventrelated synchronization/desynchronization (ERS/ERD). In this chapter, I introduce some commonly used spectral estimation methods (e.g., the periodogram, the Welch’s method, and the multitaper method) and time-frequency analysis methods (e.g., short-time Fourier transform and continuous wavelet transform). We also raise some practical issues and cautionary notes when using these methods on EEG data analysis, such as parameter tuning, visualization, and results reporting.