Third International Conference on Spoken Language Processing (ICSLP 94)
This paper describes classification methods of spoken languages based on fundamental frequency (F0) contours of speech. Speech data were taken from language learning tapes. Six languages including Japanese, Korean, Chinese, English, French and German were used. First, F0 contour was approximated by a set of polygonal lines so that the mean square error between the lines and F0 values was minimized; the optimum boundaries of the lines were determined using a dynamic programming procedure. The starting frequency, slope and duration of each line were calculated. Seventeen parameters including mean values and standard deviations of the above parameters were used for the analysis. Then parameters derived from F0pattern were analyzed by using principal component analysis. We also tried discriminant analysis of these parameters. Results show that the six languages can be classified based on these parameters. Keywords: Slope of F0 contour, Principal Component analysis, Discriminant analysis.
Bibliographic reference. Itahashi, Shuichi / Zhou, Jian Xiong / Tanaka, Kimihito (1994): "Spoken language discrimination using speech fundamental frequency", In ICSLP-1994, 1899-1902.