Variable-length unit selection using LSA-based syntactic structure cost

This paper introduces a variable-length unit selection method based on LSA-based syntactic structure for concatenative speech synthesis. First, a probabilistic context free grammar (PCFG) based parser is used to construct the syntactic structure of the input text sentence. Second, the synthesizer selects the candidate units for each node of the syntactic structure. Latent Semantic Analysis (LSA) is then adopted to estimate the syntactic cost between the target unit and the candidate units in the database. Finally, the concatenation of units with minimum cost is selected using dynamic programming algorithm. Experimental results show that variable-length unit selection based on syntactic structure outperforms the synthesizer without considering syntactic structure. Also, the LSA-based syntactic cost provides better estimation of substitution cost than that calculated only from acoustic features.