Communicative F0 generation based on impressions

Author

Lu Shao ; Greenberg, Yoko ; Sagisaka, Yoshinori

Author_Institution

Waseda Univ., Tokyo, Japan

fYear

2014

fDate

5-7 Nov. 2014

Firstpage

115

Lastpage

119

Abstract

This paper introduces our research efforts of prosody control for so-called paralinguistic information embedded in communicative speech. To specify the output prosody, we employ three-dimensional expressions extracted from 26 impressions using Multi-Dimensional Scaling. Based on a series of our previous studies showing the correlations between impressions and prosody characteristics, we propose an exact computational scheme to obtain communicative F0 using impressions given by input lexicons and the F0 pattern of corresponding reading style speech. Experimental trials have confirmed the effectiveness of the proposed calculation scheme for a set of expressions consisting of lexicons forming impressions. Finally, further advanced problems are discussed to apply the proposed scheme to other expressions.

Keywords

speech processing; speech synthesis; 3D expressions; F0 pattern; communicative F0 generation; communicative speech; computational scheme; multidimensional scaling; output prosody; paralinguistic information; prosody characteristics; prosody control; style speech; Correlation; Neural networks; Predictive models; Solid modeling; Speech; Speech synthesis; Vectors; Fundamental frequency cotrol; communicative speech synthesis; impression; para-linguistics; speech prosody;

fLanguage

English

Publisher

ieee

Conference_Titel

Cognitive Infocommunications (CogInfoCom), 2014 5th IEEE Conference on

Conference_Location

Vietri sul Mare

Type

conf

DOI

10.1109/CogInfoCom.2014.7020429

Filename

7020429