Patent
Feature parameter candidate generation apparatus and feature parameter candidate generation method
| العنوان: | Feature parameter candidate generation apparatus and feature parameter candidate generation method |
|---|---|
| Patent Number: | 8,630,823 |
| تاريخ النشر: | January 14, 2014 |
| Appl. No: | 12/740336 |
| Application Filed: | October 31, 2008 |
| مستخلص: | A feature parameter candidate generation apparatus has a storage unit that stores the values of feature parameters extracted from each of samples, an index value calculation unit that calculates an index value, which is obtained by normalizing the number of the kinds of the values of feature parameters by the number of the samples, for each of the feature parameters, an evaluation object selection unit that selects combinations of feature parameters which are objects to be evaluated, an evaluation unit that evaluates whether the uniformity of a frequency distribution of index values of the individual feature parameters for combinations of feature parameters selected as the objects to be evaluated satisfies a predetermined criterion, and a candidate determination unit that determines, as feature parameter candidates to be given to the model generation device, a combination of feature parameters that is evaluated to satisfy the predetermined criterion. |
| Inventors: | Yoneda, Mitsuhiro (Kyoto, JP); Nakajima, Hiroshi (Kyoyo, JP); Tsuchiya, Naoki (Kyoto, JP); Tasaki, Hiroshi (Kyoto, JP) |
| Assignees: | Omron Corporation (Kyoto, JP) |
| Claim: | 1. A feature parameter candidate generation apparatus for generating feature parameter candidates to be given to a model generation device that builds a model by using a plurality of feature parameters selected from given feature parameter candidates, the apparatus comprising one or more computers; and a storage unit that stores the values of a plurality of kinds of feature parameters extracted from each of a plurality of samples; wherein the one or more computers are configured with: an index value calculation unit that calculates an index value, which is obtained by normalizing the number of the kinds of the values of the feature parameter by the number of the samples, for each feature parameter of the plurality of kinds of feature parameters; an evaluation object selection unit that selects a subset of feature parameters out of the plurality of kinds of feature parameters; an evaluation unit that judges whether the subset of feature parameters selected by the evaluation object selection unit has a desired balance of the index values, by evaluating whether the uniformity of a frequency distribution of the index values satisfies a predetermined criterion; and a candidate determination unit that determines the subset of feature parameters as feature parameter candidates to be given to the model generation device, when the subset of feature parameters is judged to have the desired balance of the index values by the evaluation unit. |
| Claim: | 2. The feature parameter candidate generation apparatus as set forth in claim 1 , wherein; an update of a subset of feature parameters by the evaluation object selection unit and an evaluation of the updated subset of feature parameters by the evaluation unit are repeated so as to find out a subset of feature parameters which has a desired balance of index values. |
| Claim: | 3. The feature parameter candidate generation apparatus as set forth in claim 2 , wherein; the evaluation object selection unit updates the subset of feature parameters by adding, among the plurality of kinds of feature parameters, a feature parameter which is not included in the subset of feature parameters to the subset of feature parameters; and the evaluation object selection unit specifies a portion which indicates a minimum frequency in a frequency distribution of index values of feature parameters included in the subset of feature parameters, and selects, as a feature parameter added to the subset of feature parameters, a feature parameter which has an index value corresponding to the portion indicating the minimum frequency. |
| Claim: | 4. The feature parameter candidate generation apparatus as set forth in claim 3 , wherein; the evaluation object selection unit updates the subset of feature parameters by changing resolution of the values of at least any feature parameter among feature parameters included in the subset of feature parameters; and the evaluation object selection unit specifies a portion which indicates a maximum frequency in a frequency distribution of index values of feature parameters included in the subset of feature parameters, and changes the resolution of a feature parameter which has an index value corresponding to the portion indicating the maximum frequency. |
| Claim: | 5. The feature parameter candidate generation apparatus as set forth in claim 2 wherein; the evaluation object selection unit updates the subset of feature parameters by changing resolution of the values of at least any feature parameter among feature parameters included in the subset of feature parameters; and the evaluation object selection unit specifies a portion which indicates a maximum frequency in a frequency distribution of index values of feature parameters included in the subset of feature parameters, and changes the resolution of a feature parameter which has an index value corresponding to the portion indicating the maximum frequency. |
| Claim: | 6. A feature parameter candidate generation method for generating feature parameter candidates to be given to a model generation device that builds a model by using a plurality of feature parameters selected from given feature parameter candidates, the method wherein; a computer, which is provided with a storage unit that stores the values of a plurality of kinds of feature parameters extracted from each of a plurality of samples, executes: an index value calculation step that calculates an index value, which is obtained by normalizing the number of the kinds of the values of the feature parameter by the number of the samples, for each feature parameter of the plurality of kinds of feature parameters; an evaluation object selection step that selects a subset of feature parameters out of the plurality of kinds of feature parameters; an evaluation step that judges whether the subset of feature parameters selected by the evaluation object selection step has a desired balance of the index values, by evaluating whether uniformity of a frequency distribution of the index values satisfies a predetermined criterion; and a candidate determination step that determines the subset of feature parameters as feature parameter candidates to be given to the model generation device, when the subset of feature parameters is judged to have the desired balance of the index values by the evaluation step. |
| Claim: | 7. A non-transitory computer readable storage medium storing a feature parameter candidate generation program for generating feature parameter candidates to be given to a model generation device that builds a model by using a plurality of feature parameters selected from given feature parameter candidates, the program characterized by causing a computer, which is provided with a storage unit that stores the values of a plurality of kinds of feature parameters extracted from each of a plurality of samples, to execute: an index value calculation step that calculates an index value, which is obtained by normalizing the number of the kinds of the values of the feature parameter by the number of the samples, for each feature parameters of the plurality of kinds of feature parameters; an evaluation object selection step that selects a subset of feature parameters out of the plurality of kinds of feature parameters; an evaluation step that judges whether the subset of feature parameters selected by the evaluation object selection step has a desired balance of the index values, by evaluating whether uniformity of a frequency distribution of the index values satisfies a predetermined criterion; and a candidate determination step that determines the subset of feature parameters as feature parameter candidates to be given to the model generation device, when the subset of feature parameters is judged to have the desired balance of the index values by the evaluation step. |
| Current U.S. Class: | 703/2 |
| Patent References Cited: | 6496834 December 2002 Cereghini et al. 2003/0212678 November 2003 Bloom et al. 2003/0212692 November 2003 Campos et al. 04-084277 March 1992 9-081731 March 1997 11-126212 May 1999 |
| Assistant Examiner: | Osborne, Luke |
| Primary Examiner: | Shah, Kamini S |
| Attorney, Agent or Firm: | Foley & Lardner LLP |
| رقم الانضمام: | edspgr.08630823 |
| قاعدة البيانات: | USPTO Patent Grants |
كن أول من يترك تعليقا!