Automatically clustering similar units for unit selection in speech synthesis

By Alan W. Black, Paul A. Taylor. 1997

This paper describes a new method for synthesizing speech by concatenating sub-word units from a database of labelled speech. A large unit inventory is created by automatically clustering units of the same phone class based on their phonetic and prosodic context. The appropriate cluster is then selected for a target unit offering a small set of candidate units. An optimal path is found through the candidate units based on their distance from the cluster center and an acoustically based join cost. Details of the method and justification are presented. The results of experiments using two different databases are given, optimising various parameters within the system. Also a comparison with other existing selection based synthesis techniques is given showing the advantages this method has over existing ones. The method is implemented within a full text-to-speech system offering efficient natural sounding speech synthesis.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Neural Speech Synthesis with Transformer Network (2018, methodology): “Concatenative TTS systems (Hunt and Black 1996; Black and Taylor 1997) and stitch them by some algorithms such as Viterbi (Viterbi 1967) followed by signal process methods (Charpentier and Stella 1986; Verhelst and Roelands 1993) to generate new waves.”
Search Result Clustering in Collaborative Sound Collections (2020, background): “from a feature space, on which a numerical distance or similarity is calculated. When clustering audio content, the features and distance measures are often chosen carefully for a given specific task [2, 26, 30]. However in the context of large online collections, the content is very diverse, contai”
Close to Human Quality TTS with Transformer (2018, unknown): “Concatenative TTS systems (Hunt and Black 1996; Black and Taylor 1997) Traditional speech synthesis methods require extensive domain expertise and may contain brittle design choices.”
Text-to-Speech Synthesis (2009, background): “Partly because of its institutional nature, where visiting researchers would come for a few months to a few years and then leave to another organisation, ATR spawned a whole “family tree” of synthesis systems, including the Festival work at Edinbu rgh university [49] and the AT&T Next Gen system [42”
DurIAN: Duration Informed Attention Network For Multimodal Synthesis (2019, methodology): “Traditional speech synthesis approaches, including concatenative methods [1, 2] and statistic parametric systems [3, 4, 5], are all based on acoustic feature analysis and synthesis.”
An HMM-based speech synthesis system applied to English (2002, methodology): “To avoid this, a clustering-based scheme was also developed [17].”
Statistical Parametric Speech Synthesis (2007, methodology): “2, uses a clustering method that allows the target cost to effectively be pre-calculated (Donovan and Woodland, 1995; Black and Taylor, 1997).”
Flow-TTS: A Non-Autoregressive Network for Text to Speech Based on Flow (2020, methodology): “Concatenative TTS [1, 2] and statistical parametric TTS [3, 4] are the two most successful TTS techniques in the past decades.”
DurIAN: Duration Informed Attention Network for Speech Synthesis (2020, methodology): “Traditional speech synthesis approaches, including concatenative methods [1, 2] and statistic parametric systems [3, 4, 5], are all based on acoustic feature analysis and synthesis.”
Natural TTS Synthesis by Conditioning Wavenet on MEL Spectrogram Predictions (2017, background): “Concatenative synthesis with unit selection, the process of stitching small units of pre-recorded waveforms together [2, 3] was the state-of-the-art for many years.”
Forward–Backward Decoding Sequence for Regularizing End-to-End TTS (2019, background): “[17]–[20] are proposed to simplify traditional TTS pipeline [3], [6], [11], [12] with a single neural network.”