摘要 |
PROBLEM TO BE SOLVED: To automatically exclude a section in which a noise component is dominant, to extract a section of high similarity and to evaluate the similarity even when a signal and a noise are non-stationary. SOLUTION: In two signals of which similarity is to be compared, a signal f(x) is divided into small areas, one small area is selected and a correlation with a signal g(x) is calculated using a correlation method. Then, a parallel distance, a magnification ratio and a similarity in a area where the similarity of the square of a correlation value becomes maximum are sought, and for the position represented by the parallel distance and the magnification ratio, a value based on the similarity is totalized. Similar processing is performed for every small area, and, at a peak where the totalized value of the similarity becomes maximum, its magnitude is compared with a threshold to evaluate the similarity. In addition, a small area that has voted for the peak can be extracted. |