Comparative analysis between SROL - Romanian database and Emo - German database

Author

Feraru, Silvia Monica ; Zbancioc, Marius Dan

Author_Institution

Inst. of Comput. Sci., Iaşi, Romania

fYear

2015

Firstpage

1

Lastpage

4

Abstract

In this paper, we compared the recognition rate obtained with our instruments for the emotional states (happiness, fury, sadness and neutral tone) from the SROL and from the Emo databases. The main goal of this research is to validate the instruments used for the extracting of the features vectors applied for the automatic recognition of the emotions and the results presented in our previously papers, for the Romanian language. The global recognition percents of the emotions is around 90% using Weighted K-Nearest Neighbors - WKNN classifier for the SROL Romanian database, respectively 86% for Emo German database. The recognition rates are close to the other studies from the literatures, for German database.

Keywords

audio databases; emotion recognition; pattern classification; speech processing; Emo German database; SROL Romanian database; WKNN classifier; automatic emotion recognition; emotional states; features vectors; global recognition percents; recognition rate; weighted k-nearest neighbors; Databases; Emotion recognition; Feature extraction; Hidden Markov models; Mel frequency cepstral coefficient; Speech; Support vector machines;

fLanguage

English

Publisher

ieee

Conference_Titel

Signals, Circuits and Systems (ISSCS), 2015 International Symposium on

Conference_Location

Iasi

Print_ISBN

978-1-4673-7487-3

Type

conf

DOI

10.1109/ISSCS.2015.7204015

Filename

7204015