| 摘要 |
<P>PROBLEM TO BE SOLVED: To provide a similarity evaluation device that evaluate similarities of time-series data to be evaluated to model time-series data sequentially by individual pattern sections. <P>SOLUTION: The model time-series data is divided into sections Y<SB POS="POST">i</SB>(i=1, ..., n) of an individual pattern at predetermined places in time series in advance, and feature quantity time series y<SB POS="POST">i</SB>(i=1, ..., n) by the sections are extracted. In order to evaluate similarities of sequentially input time-series data to be evaluated sequentially by individual pattern sections X<SB POS="POST">i</SB>(i=1, ..., n) to be evaluated which have the same intervals with the predetermined divisions, the similarity evaluation device 1 includes a feature quantity extraction part 2 which extracts a feature quantity time series x<SB POS="POST">i</SB>of the same kind with the time series extracted from the model time-series data in advance, and a similarity calculation part 4 which finds similarities between the feature quantity time series x<SB POS="POST">i</SB>of the individual sections X<SB POS="POST">i</SB>and the feature quantity time series y<SB POS="POST">i</SB>of the model individual pattern sections Y<SB POS="POST">i</SB>. Main component scores may be used as the feature quantities. <P>COPYRIGHT: (C)2013,JPO&INPIT |
