Time-frequency and time-scale distributions provide a powerful tool for non-stationary signal analysis. Unlike conventional spectral methods, they reveal the time-varying spectral content of one-dimensional signals by mapping them into a two-dimensional time-frequency domain. Substantial theoretical work has been carried out in this direction and has yielded many different classes of time-frequency distributions, parametric or otherwise, in which optimal solutions for a given signal or task can be selected. My current research interests involve the development of an operational framework for testing stationarity relatively to an observation scale, in both stochastic and deterministic contexts, and pattern recognition algorithms in the time-frequency domain based on machine learning techniques.