Maximum Likelihood and Maximum a Posteriori Adaptation for Distributed Speaker Recognition Systems