Output details
11 - Computer Science and Informatics
University of Edinburgh
HMM-Based Speech Synthesis Utilizing Glottal Inverse Filtering
<22>Originality: First paper to describe a HMM-based text-to-speech synthesizer that utilizes glottal inverse filtering techniques for accurately modelling glottal components of speech and for generating higher-quality synthetic speech.
Significance: Detailed experiments show that the proposed system is capable of generating natural sounding speech, and the quality is better than a state-of-the-art HMM-based speech synthesis system based on widely-used vocoder technique called STRAIGHT – this is viewed as a very remarkable result in the speech synthesis field.
Rigour: A probabilistic acoustic model was carefully developed to model characteristics of glottal speech waveforms. The proposed approach was evaluated using a Finnish corpus.