Time–Frequency Feature Analysis of Speech Signals Using Wavelet Scalograms and Gabor Transform

Speech signals are inherently non-stationary and require effective time–frequency analysis techniques to capture their dynamic characteristics. This study investigates the extraction and analysis of time–frequency features from speech signals using wavelet scalograms and the Gabor transform. The wavelet scalogram provides a multi-resolution representation of the signal, enabling detailed analysis of both high- and low-frequency components over time. In contrast, the Gabor transform offers a localized time–frequency representation using Gaussian-windowed sinusoids, which is well suited for analyzing spectral structures in speech. In this work, speech signals are processed using both approaches to obtain their respective time–frequency coefficients. The resulting representations are then analyzed to examine differences in resolution, feature localization, and their effectiveness in capturing speech characteristics. Experimental results demonstrate that the wavelet scalogram provides better adaptability to transient features, while the Gabor transform offers consistent resolution across the time–frequency plane. The comparative analysis highlights the strengths of each method and their potential applications in speech analysis, feature extraction, and signal characterization.