Time-Frequency Analysis of Non-Stationary Signals Based on Short-Time Fourier Transform

Time-frequency analysis based on the Short-Time Fourier Transform (STFT) is a powerful method for studying the characteristics of non-stationary signals in both time and frequency domains. The STFT technique decomposes signals into time-frequency representations by applying Fourier transforms to windowed segments of the signal, effectively revealing how signal components evolve across different time instances and frequency bands. In practical implementation, the STFT algorithm typically involves sliding a window function (such as Hanning, Hamming, or Gaussian windows) along the signal timeline and computing the Fourier transform for each windowed segment. Key parameters include window size, which controls the trade-off between time and frequency resolution, and overlap between consecutive windows to reduce spectral leakage. The resulting spectrogram provides a visual representation of signal energy distribution across time and frequency. Through time-frequency analysis of non-stationary signals, researchers can gain deeper insights into temporal and spectral characteristics, as well as their interrelationships. This approach finds extensive applications across multiple domains including speech processing (for analyzing time-varying speech features), image processing (for texture analysis), biomedical engineering (for EEG and ECG signal analysis), and vibration analysis in mechanical systems. The MATLAB implementation typically utilizes functions like spectrogram() or stft() with customizable parameters for window type, size, and overlap percentage. Therefore, time-frequency analysis of non-stationary signals based on STFT represents a significant and valuable research area, with practical code implementations involving careful parameter selection and optimization for specific application requirements.