| 摘要 |
PROBLEM TO BE SOLVED: To approximate waveform data to an appropriate non-uniform B spline curve without requiring unnecessary parameter tuning by a user.SOLUTION: An information processor selects a plurality of section width trains of the intermediate knots in a non-uniform B splice curve which approximate waveform data based on a multivariate normal distribution. NOT vector and the optimum control point train on each vector are decided based on each of the selected section width train. The information processor repeats the selection of the section width train, the decision of the not vector and the optimum control point train, and its estimation until the estimation in which an object function becomes minimum is acquired with respect to the non-uniform B spline curve regulated regarding each combination of the NOT vector and the optimum control point train corresponding to the NOT vector. During this period, the average and dispersion of the multivariate normal distribution among the plurality of selected section width trains are updated based on a prescribed number of section width trains selected in the ascending order of the approximate difference, and after that, the repeated operations of the next generation are performed using the multivariate normal distribution after updated. |
