An Overview of Speech Synthesis Technology

Speech is the most natural and convenient approach of communication and speech synthesis technology is a kind of import application in Human-machine interaction system. This paper gives a comprehensive overview of Text-to-Speech (TTS) synthesis technology. The two basic parts of speech synthesis technology are natural language processing (NLP) and digital signal processing (DSP). To the part of NLP, some important steps are pre-processing, morphological analysis, contextual analysis, syntactic-prosodic analysis, phonetization and prosody generation. To the part of DSP, two types of synthesis methods are rule-driven methods and data-driven methods. Some important synthesis approaches of DSP such as articulatory synthesis, formant synthesis, concatenative synthesis, unit selection synthesis, HMM synthesis and DNN synthesis are introduced. Finally, these approaches of speech synthesis are compared briefly. The technical trends of TTS and some hot spots of its applications in the future are discussed.