Pattern recognition method and pattern recognition apparatus
A pattern recognition apparatus that recognizes a data attribute of input data calculates correlation values of feature quantities of corresponding local patterns between the input data and dictionary data for each of a plurality of dictionary data prepared for each data attribute, combines, for each data attribute, the calculated correlation values of local patterns of each dictionary datum to acquire a set of correlation values of each data attribute, integrates correlation values included in each set of correlation values of each data attribute to calculate a similarity of the input data for each data attribute, and identifies the data attribute of the input data based on the calculated similarity.
1. A pattern recognition apparatus that recognizes a class of an input object pattern, the apparatus comprising:
circuitry functioning as:
a correlation value calculation unit configured to calculate a correlation value between a feature quantity of each local area in the input object pattern and a feature quantity of a corresponding local area in each of registered object patterns in a dictionary in which at least one object pattern has been registered for each of predetermined classes and a plurality of object patterns have been registered for at least one of the predetermined classes;
a combining unit configured to combine, for each class of the predetermined classes, the correlation values calculated for respective registered object patterns which belong to that class with respect to each local area in the input object pattern;
an obtaining unit configured to obtain an integrated similarity of the input object pattern for each class of the predetermined classes by integrating the combined correlation values of a plurality of local areas for that class; and
an identification unit configured to identify a class of the input object pattern based on the integrated similarity for each class obtained by said obtaining unit.
2. The apparatus according to claim 1 , further comprising:
an estimation unit configured to estimate a variation pattern in the input object pattern; and
a determination unit configured to determine a local pattern used for each dictionary datum based on the variation pattern of the input object pattern and a variation pattern of each dictionary datum, wherein
said combining unit calculates the similarity of the input object pattern for each class by integrating the set of correlation values of each dictionary datum corresponding to the local pattern to be used to acquire combined correlation values.
3. The apparatus according to claim 1 , wherein
the combined correlation value is an average value of only the correlation values over the threshold, a weighted average value of the correlation values, or a maximum value of the correlation values.
4. The apparatus according to claim 1 , wherein
said obtaining unit calculates an average value of the combined correlation values over the threshold, a weighted average value of the combined correlation values, or a maximum value of the combined correlation values for each set of combined correlation values of each class as the integrated similarity for the class.
5. The apparatus according to claim 2 , wherein
the variation pattern is at least one of variations in face orientation of a subject included in the object pattern, variations in illumination, variations in expression of the face, and occlusions of the face.
6. The apparatus according to claim 1 , wherein
the recognized class is individual identification information.
7. A pattern recognition method by a pattern recognition apparatus that recognizes a class of an input object pattern, the method comprising:
a correlation value calculation step of calculating a correlation value between a feature quantity of each local area in the input object pattern and a feature quantity of a corresponding local area in each of registered object patterns in a dictionary in which at least one object pattern has been registered for each of predetermined classes and a plurality of object patterns have been registered for at least one of the predetermined classes;
a combining step of combining, for each class of the predetermined classes, the correlation values calculated for respective registered object patterns which belong to that class with respect to each local area in the input object pattern;
an obtaining step of obtaining an integrated similarity of the input object pattern for each class of the predetermined classes by integrating the combined correlation values of a plurality of local areas for that class; and
an identification step of identifying a class of the input object pattern based on the integrated similarity for each class obtained in said obtaining step.
8. A non-transitory computer-readable storage medium storing a computer program causing a computer to execute the steps of the pattern recognition method according to claim 7 .
9. The apparatus according to claim 1 , wherein the circuitry comprises a processor.
10. The apparatus according to claim 1 , wherein the circuitry comprises dedicated arithmetic circuits or firmware.
11. The apparatus according to claim 3 , wherein
said obtaining unit calculates an average value of the combined correlation values over the threshold, a weighted average value of the combined correlation values, or a maximum value of the combined correlation values for each set of combined correlation values of each class as the integrated similarity for the class.