Speaker adaptation using context clustering decision tree for HMM-based speech synthesis