Robust time-frequency distributions

Study of the additive Gaussian noise influence on time-frequency (TF) distributions is an important issue [Article 9.1]. However, in many practical applications, especially in communications, signals are disturbed by a kind of impulse noise. These noises are commonly modeled by heavy-tailed (long-tailed) probability density functions (pdfs) [6]. It is well known fact that the conventional TF distributions are quite sensitive to this kind of noise, which is able to destroy sensitive signal information. The minimax Huber M-estimates [5] can be applied in order to design the periodogram and TF distributions, robust with respect to the impulse noise. For nonstationary signals, the robust TF distributions are developed as an extension of the robust M estimation approach.