Detection of Time - Frequency Components of Signals : Bayes and focus SNR

The Bayesian time-frequency detector operating on spectrogram of non stationary signals is studied. As direct evaluation of the likelihood ratio is impossible, an a priori user parameter called focus is introduced. It is defined as a local time-frequency signal to noise ratio at which the detection is tuned to be optimal. Leading to a unique detection threshold, the parameter introduced is equivalent to the probability of false alarm used in the Neyman-Pearson detection strategy. However, we expect the formulation in terms of local signal to noise ratio to be of intuitive and practical interest.