The Breath Segment in Expressive Speech

Computational Linguistics and Chinese Language Processing Vol. 12, No. 1, March 2007, pp. 17-32 17  The Association for Computational Linguistics and Chinese Language Processing [Received August 12, 2006; Revised October 22, 2006; Accepted November 1, 2006] The Breath Segment in Expressive Speech Chu Yuan∗, and Aijun Li∗ Abstract This paper, based on a selected one hour of expressive speech, is a pilot study on how to use breath segments to get more natural and expressive speech. It mainly deals with the status of when the breath segments occur and how the acoustic features are affected by the speaker’s emotional states in terms of valence and activation. Statistical analysis is made to investigate the relationship between the length and intensity of the breath segments and the two state parameters. Finally, a perceptual experiment is conducted by employing the analysis results to synthesized speech, the results of which demonstrate that breath segment insertion can help improve the expressiveness and naturalness of the synthesized speech. Keywords: Breath Segment, Expressive Speech, Emotion, Valence, Activation 1. Introduction In the current speech synthesis and recognition systems, some characteristics of spontaneous speech are treated as noise, such as disfluent utterances, repeated sounds, filled pauses, salient breaths and coughs. In corpus collection for speech synthesis and recognition systems, the speaking style of the speakers is always strictly controlled and the speaker is usually required to give a “canonical pronunciation” to decrease the speaking noise as much as possible. However, in recent study, researchers have begun to pay more attention to the non-verbal information in natural speech, especially the paralinguistic and physiological information. They have focused on how to use these types of information to improve the naturalness and expressiveness of emotion and attitude in synthesized speech, so that the speaker ’s intention can be better understood during verbal communication. In 1989, Cahn compiled a simple feeling editor based on the phonetic characteristics of emotion [Cahn 1990]. Vroomen, Collier and Mozziconacci examined the duration and intonation of emotional speech and proposed that emotions can be expressed accurately by manipulating pitch and duration based on rules. This conclusion showed that, in emotional ∗ Institute of Linguistics, Chinese Academy of Social Sciences, No. 5 Jianguomennei Dajie, Beijing, 100732 China E-mail: Yuanchu8341@gmail.com; liaj@cass.org.cn 18 Chu Yuan and Aijun Li speech, duration and intonation can be employed to observe the speakers ’ attitude [Vroomen et al. 1993]. In 1998, Campbell found that if one compares the same content in different forms, for example, a news item in its read form, its formal spoken or broadcast form, and its informal conversational form, differences are obvious not only in lexis, word-order, chunking, and prominence relations, but also in the mood of the speaker and in the tone of the voice [Campbell 1998]. In 2000, the International Workshop on Speech and Emotion of ISCA (held in Ireland) invited, for the first time, researchers who were devoted to the study of emotion and speech. Before this conference, many researchers had begun to investigate the voice quality, …

Content truncated to 500 words