Time-frequency analysis of Interference in complex signals – TIPHAINE

Multicomponent signals are widely encountered in various real-life applications such as audio (music, speech), physiological monitoring (e.g., electrocardiogram, phonocardiogram, electroencephalogram), astronomy (e.g., gravitational waves) or echolocation (used by bats and marine mammals). These signals are typically modeled as the superimposition of amplitude/frequency modulated modes. Disentangling these constituent modes in the time-frequency plane has been the subject of intense research for decades.
To that aim, the first major challenge consists in the design of a time-frequency representation enabling mode separation in the time-frequency plane, which may be very tough either when some noise is present, when the modes are close or overlapping in the time-frequency plane, or when both conditions occur. A second challenge is to sharpen the time-frequency representation of signals to improve their readability. Both of these challenges are deeply interconnected, since most techniques for improving the readability of time-frequency representations rely on effective mode separation in the TF plane. In the present project, our goal is thus to work jointly on these two aspects, by building new adaptive sharp time-frequency representations, bearing in mind the potential presence of close or crossing modes, and paying a particular attention to the computational cost of the proposed techniques. In these new developments, accurately localizing and characterizing the interference in the time-frequency plane will be of paramount importance.