On the Estimation of the Short-Time Phase in Speech Enhancement Systems