METHOD, DEVICE AND PROGRAM FOR NON-STEADY TIME-SERIES DATA CLASSIFICATION, AND RECORDING MEDIUM RECORDED WITH THIS PROGRAM

摘要

PROBLEM TO BE SOLVED: To classify non-steady time-series data including a missing value like an F0 pattern based upon objective measures. SOLUTION: The time-series data y(s)tto be classified are read in (step S1) and a polynomial regression model is applied thereto to generate an estimation function T^(s)(t)=(1, t, t2, ..., tp-1)β^(s)LSEof a trend function showing variation trends of the time-series data y(s)tfor each sample (s) (step S2). Here, the order (p) is so determined that an information amount standard AIC(s)is minimized, and the regression coefficient estimatedβ^(s)LSE is calculated by a least-squares method. Then a decision function is generated by two-group decision analysis from this estimation function T^(s)(step S3) and used to perform cluster analysis of the estimation function T^(s)(t) (step S4) and its classification result is outputted (step S5). COPYRIGHT: (C)2006,JPO&NCIPI