Adaptive Kalman filter for speech enhancement from colored noise

In the framework of speech enhancement we propose a new approach for signal recovering in colored noise based on adaptive Kalman filter. The approaches proposed in the past, in this context, operate in two steps: they first estimate the noises variances and the parameters of the signal and noise models and secondly estimate the speech signal. In this paper we propose a new parameters estimation method based on the EM (Expectation-Maximisation) algorithm.