engleski

A Comparison of Two Approaches to Bilingual HMM- Based Speech Synthesis

We compare the performance of two approaches when using cross-lingual data from different speakers to build bilingual speech synthesis systems capable of producing speech with the same speaker identity. One approach treats data from both languages as monolingual, by labeling all data with a manually joined phoneme set. Speaker independent voice is trained using the joined data, and adapted to the target speaker using the CMLLR adaptation. In the second approach, speaker independent voices are trained for each language separately. State mapping between these voices is derived automatically from minimum Kullback–Leibler divergence between state distributions. The mapping is used to apply the adaptation transformations calculated within one language across languages to the other speaker independent voice. We evaluate the quality of speech on MOS scale and similarity of synthesized speech characteristics to the target speaker using DMOS on the example of Croatian-Slovene language pair.

bilingual; HMM; speech synthesis; phoneme mapping; state mapping; speaker adaptation; Kullback-Leibler divergence

nije evidentirano