Modified quadratic discrimination function for non-normal distribution and its application in speech emotion recognition

Author

Yan, Zhao ; Li, Zhao ; Cairong, Zou ; Yinhua, Yu ; Chengwei, Huang ; Qingyun, Wang

Author_Institution

Sch. of Inf. Sci. & Technol., Southeast Univ., Nanjing

fYear

2008

fDate

Nov. 30 2008-Dec. 3 2008

Firstpage

213

Lastpage

216

Abstract

The quadratic discrimination function (QDF) is a common method widely used in pattern recognition. It is optimal when the sampled features are normal with known means and covariance matrices. In this paper, a new transform is proposed and a method named modified QDF (MQDF) is deduced aim to improve the non-normality of the features during the speech emotion recognition, both prosodic and voice quality features are considered in the experiment. The results show that our method is effective and robust.

Keywords

covariance matrices; speech recognition; covariance matrices; modified quadratic discrimination function; nonnormal distribution; pattern recognition; prosodic features; speech emotion recognition; voice quality features; Bayesian methods; Covariance matrix; Emotion recognition; Error probability; Estimation error; Gaussian distribution; Information science; Pattern recognition; Robustness; Speech recognition; Modified quadratic discrimination function (MQDF); normalization; speech emotion recognition;

fLanguage

English

Publisher

ieee

Conference_Titel

Circuits and Systems, 2008. APCCAS 2008. IEEE Asia Pacific Conference on

Conference_Location

Macao

Print_ISBN

978-1-4244-2341-5

Electronic_ISBN

978-1-4244-2342-2

Type

conf

DOI

10.1109/APCCAS.2008.4745998

Filename

4745998