Variable-Length Unit Selection in TTS Using Structural Syntactic Cost

• Published in: IEEE transactions on audio, speech, and language processing Vol. 15; no. 4; pp. 1227 - 1235
• Main Authors: Chung-Hsien Wu, Hsia, C.-C., Jiun-Fu Chen, Jhing-Fa Wang
• Format: Journal Article
• Language: English
• Published: Piscataway, NJ IEEE 01.05.2007; Institute of Electrical and Electronics Engineers
• Subjects: Algorithms; Applied sciences; Computer science; Concatenated codes; Costs; Dynamic programming; Exact sciences and technology; Heuristic algorithms; Humans; Information, signal and communications theory; Latent semantic analysis (LSA); Mathematical analysis; probabilistic context-free grammar (PCFG); Semantics; Signal processing; Spatial databases; Speech; Speech processing; Speech recognition; Speech synthesis; Statistical analysis; syntactic structure; Synthesizers; Telecommunications and information theory; variable-length unit selection; Vectors (mathematics); Performance evaluation; Speech analysis; Probabilistic approach; Semantic analysis; syntactic structure; Latent semantic analysis (LSA); Verbal perception; Speech synthesis; Algorithm; Statistical test; Linguistic analysis; Speech recognition; Database; Context free grammar; Economic aspect; Dynamic programming; variable-length unit selection; Cost analysis; Syntax; Speech processing; probabilistic context-free grammar (PCFG)
• ISSN: 1558-7916; 1558-7924
• DOI: 10.1109/TASL.2006.889752

This paper presents a variable-length unit selection scheme based on syntactic cost to select text-to-speech (TTS) synthesis units. The syntactic structure of a sentence is derived from a probabilistic context-free grammar (PCFG), and represented as a syntactic vector. The syntactic difference between target and candidate units (words or phrases) is estimated by the cosine measure with the inside probability of PCFG acting as a weight. Latent semantic analysis (LSA) is applied to reduce the dimensionality of the syntactic vectors. The dynamic programming algorithm is adopted to obtain a concatenated unit sequence with minimum cost. A syntactic property-rich speech database is designed and collected as the unit inventory. Several experiments with statistical testing are conducted to assess the quality of the synthetic speech as perceived by human subjects. The proposed method outperforms the synthesizer without considering syntactic property. The structural syntax estimates the substitution cost better than the acoustic features alone