Some contributions to joint optimal filtering and parameter estimation with application to monaural speech separation