Title :
Separability and recognition of emotion states in multilingual speech
Author :
Xiaoqing, Jiang ; Lan, Tian ; Min, Han
Author_Institution :
Sch. of Inf. Sci. & Eng., Shandong Univ., China
Abstract :
In this paper we analyze the separability and recognition of emotion states in multilingual speech signals statistically. Prosodic features such as pitch, energy and time parameters are extracted and the separability is discussed based on the statistical results in a two-dimension method. Principal component analysis (PCA) is then used to recognize emotion states and achieved satisfying results, in which mean recognition rate is 72.26% and the highest recognition rate is 89%. The results show that language factors do not affect features of prosodic variation of some given emotion obviously and basic emotions can be recognized roughly from speech signals using prosodic parameters.
Keywords :
emotion recognition; feature extraction; human computer interaction; linguistics; principal component analysis; speech recognition; PCA; emotion state recognition; emotion state separability; human-computer interaction; language factors; multilingual speech signals; principal component analysis; prosodic features; prosodic parameters; prosodic variation; recognition rate; speech energy; speech pitch; speech time parameters; statistical results; Cepstral analysis; Cepstrum; Emotion recognition; Natural languages; Principal component analysis; Speech analysis; Speech enhancement; Speech recognition; Speech synthesis; Testing;
Conference_Titel :
Communications, Circuits and Systems, 2005. Proceedings. 2005 International Conference on
Print_ISBN :
0-7803-9015-6
DOI :
10.1109/ICCCAS.2005.1495245
