Unsupervised cross-lingual speaker adaptation for HMM-based speech synthesis using two-pass decision tree construction | IEEE Conference Publication | IEEE Xplore