Attribute estimation system, age estimation system, gender estimation system, age and gender estimation system and attribute estimation method

1. An age estimation system, having a processor, and a memory storing instructions to be executed by the processor, that estimates an age based on at least one piece of input data, wherein the ages are separated into a plurality of classes, each indicating a predetermined age group, the system comprising:

a first estimation unit which includes at least one first feature extraction unit which extracts at least one first feature from said input data, and which estimates said age as a discrete quantity based on said first feature;

a second estimation unit which includes at least one second feature extraction unit which extracts at least one second feature from said input data, and which estimates said age as a continuous quantity based on said second feature;

a score-generation unit which generates a score, which is a function indicating a relation between respective estimated results of said first estimation unit and estimated results of said second estimation unit, and age;

an integration unit which outputs an estimated age based on a score obtained by integrating a score of an estimated result of said first estimation unit and a score of an estimated result of said second estimation unit.

2. An age estimation system according to claim 1 , wherein said score-generation unit generates the score of the estimated result of said first estimation unit such that said estimated result becomes a fixed value in any one of the plurality of classes.

3. An age estimation system according to claim 1 , wherein said score-generation unit generates the score of the estimated result of said first estimation unit such that said estimated result becomes biggest at a median in any one of the plurality of classes, each indicating a predetermined age group, and decreases proportionally to a difference with said median.

4. An age estimation system according to claim 1 , wherein said score-generation unit generates the score of the estimated result of said first estimation unit such that said estimated result is of a shape of normal distribution centering on a median in any one of the plurality of classes, each indicating a predetermined age group.

5. An age estimation system according to claim 1 , wherein said score-generation unit generates the score of the estimated result of said first estimation unit as distribution of criterion data estimated for any one of the plurality of classes, each indicating a predetermined age group, based on a result which is obtained by performing a reverse lookup of criterion data which said first estimation unit has been used for learning.

6. An age estimation system according to claim 1 , wherein said score-generation unit generates the score of the estimated result of said first estimation unit such that said estimated result becomes a fixed value within each of the plurality of classes indicates a predetermined age group based on a probability corresponding to each of the plurality of classes.

7. An age estimation system according to claim 1 , wherein said score-generation unit generates for each of the plurality of classes, indicating a predetermined age group, the score of the estimated result of said first estimation unit such that said estimated result becomes biggest at a median in each of the plurality of classes, and decreases proportionally to a difference with said median based on a probability corresponding to each class, and wherein said score-generation unit accumulates the scores to generating a score.

8. An age estimation system according to claim 1 , wherein said score-generation unit generates for each of the plurality of classes, indicating a predetermined age group, a score of the estimated result of said first estimation unit such that said estimated result is of a shape of normal distribution centering on a median in the each of the plurality of classes based on a probability corresponding to each of the plurality of classes; and wherein said score-generation unit accumulates the scores to generating a score.

9. An age estimation system according to claim 1 , wherein said score-generation unit generates the score of the estimated result of said first estimation unit as distribution of criterion data estimated for each of the plurality of classes, each indicating a predetermined age group, according to a probability corresponding to each of the plurality of classes based on a result which is obtained by performing a reverse lookup of criterion data which said first estimation unit has been used for learning.

10. An age estimation system according to claim 1 , wherein said score-generation unit generates the score of the estimated result of said second estimation unit as a fixed value within a predetermined age group which takes an estimated result of said second estimation unit as a median.

11. An age estimation system according to claim 1 , wherein said score-generation unit generates the score of the estimated result of said second estimation unit such that said estimated result becomes biggest at a point of an estimated result of said second estimation unit, and decreases proportionally to a difference with said estimated result of said second estimation unit.

12. An age estimation system according to claim 1 , wherein said score-generation unit generates the score of the estimated result of said second estimation unit as a form of normal distribution centering on an estimated result of said second estimation unit.

13. An age estimation system according to claim 1 , wherein said score-generation unit generates the score of the estimated result of said second estimation unit as distribution of criterion data estimated for a certain age based on a result which is obtained by performing a reverse lookup of criterion data which said second estimation unit has been used for learning.

14. An age estimation system according to claim 1 , wherein at least one piece of said input data is image data and an age of a person photographed in said image is estimated.

15. An age estimation system according to claim 1 , wherein a plurality of images are included as said input data and an age of a person who has been photographed in common in said plurality of images is estimated.

16. An age estimation system according to claim 1 , wherein at least one piece of said input data is voice data.

17. An age estimation system according to claim 1 , wherein said integration unit outputs a peak value in an integration result of a score of an estimated result of said first estimation unit and a score of an estimated result of said second estimation unit as an estimated age.

18. An age estimation system according to claim 1 , wherein said integration unit outputs a class indicating a predetermined age group to which a peak value in an integration result of a score of an estimated result of said first estimation unit and a score of an estimated result of said second estimation unit belongs as an estimated age.

19. An age estimation system according to claim 1 , wherein said integration unit performs integral calculation of an integration result of a score of an estimated result of said first estimation unit and a score of an estimated result of said second estimation unit for each of the plurality of classes, indicating a predetermined age group, and outputs a class in which an value obtained by performing said integral calculation takes a maximum value as an estimated age.

20. An age estimation method for estimating an age based on at least one piece of input data, comprising;

extracting at least one first feature from said input data;

estimating said age as a discrete quantity based on the at least one first feature;

extracting at least one second feature from said input data;

estimating said age as a continuous quantity based on the at least one second feature

generating a score which is a function indicates a relation between respective results estimated as the discrete quantity and results estimated as the continuous quantity, and age; and

outputting an estimated age based on a score obtained by integrating a score of a result estimated as the discrete quantity and a score of a result estimated as the continuous quantity.