Classification of Spoken Languages by Fundamental Frequency

SASANUMA Mitsuru, ITAHASHI Shuichi

Bibliographic Information

Other Title

基本周波数による多言語音声の分類

Description

For the purpose of characterizing language prosodic features, fundamental frequency contours had been approximated with polygonal lines so far. This paper suggests a new model which combines exponential functions and lines. The model was determined using dyanamic programing procedure. The coefficients, relative starting frequency and duration of each lines and exponential functions were calculated and used for principal component analysis and discriminant analysis. Results of principal component analysis showed that approximation with the new model was effective for classification of spoken languages. We obtained better discriminant score using approximation with exponential function than lines only. Six language was selected from a CD-ROM that includes 21 language data. Four speaker was used for the analysis.

Journal

IEICE technical report. Speech

IEICE technical report. Speech 96 (319), 13-20, 1996-10-18

The Institute of Electronics, Information and Communication Engineers

Keywords

Details 詳細情報について

CRID: 1573668927129955712

NII Article ID: 110003296261

NII Book ID: AN10013221

Text Lang: ja

Data Source

CiNii Articles

Export

Report a problem