Information processing apparatus and information processing method

• US 7,881,524 B2
• Filed: 04/04/2006
• Issued: 02/01/2011
• Est. Priority Date: 04/08/2005
• Status: Active Grant

• Current Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Original Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Inventors: Sato, Hiroshi, Kato, Masami, Mitarai, Yusuke, Matsugu, Masakazu, Kaneda, Yuji, Mori, Katsuhiko
• Primary Examiner(s): Mehta; Bhavesh M
• Assistant Examiner(s): Rosario; Dennis
• Application Number:
  US11/396,701
• Publication Number:
  US 20060228005A1
• Time in Patent Office: 1,764 Days
• Field of Search: 382/115, 382/156, 706/20
• US Class Current: 382/156
• CPC Class Codes: G06V 40/10 Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands

Abstract

A plurality of kinds of feature amounts are collected from image information and voice information on a person existing in a space, valid values of the collected feature amounts are calculated, feature amounts to be used for personal recognition are determined in the collected feature amounts on the basis of the calculated valid values, and personal recognition is performed by using the determined feature amounts.

Claims

1. An information processing apparatus, the apparatus comprising:

a plurality of sensing means for sensing a plurality of types of data from a space in which a person exists;

detection means for detecting a plurality of kinds of feature amounts from the plurality of types of data sensed by said plurality of sensing means;

calculation means for calculating respective valid values of the plurality of kinds of feature amounts detected by said detection means;

selection means for selecting at least one feature which has a relatively high valid value from among the plurality of kinds of feature amounts, on the basis of the respective valid values calculated by said calculation means; and

identification means for identifying the person by using at least the selected one feature amount with the relatively high valid value,

wherein said calculation means includes;

means for performing processing for obtaining normalized output level Vi of modules for detecting the feature amounts with respect to all classes i in a final layer of a hierarchical neural network;

means for calculating a sum total S=V₁+V₂+, . . . V_N of output levels in all the classes; and

means for calculating a valid value Yi of the feature amount in class i by computing Yi=ki×

Vi/S where ki represents the appearance frequency in class i.

2. The information processing apparatus according to claim 1, wherein each of said plurality of sensing means senses one or more of image information of the space, voice information in the space and living body information on the person existing in the space.

3. The information processing apparatus according to claim 2, wherein each of said plurality of sensing means senses image information of the space and said detection means detects an image feature amount relating to an appearance of the person or an action of the person.

4. The information processing apparatus according to claim 2, wherein each of said plurality of sensing means senses voice information in the space and said detection means detects a voice feature amount relating to a nonlinguistic voice.

5. The information processing apparatus according to claim 2, wherein each of said plurality of sensing means senses at least one of an iris pattern, a fingerprint pattern, a vein pattern, gene arrangement information and smell as living body information on the person existing in the space.

6. The information processing apparatus according to claim 1, wherein said selection means selects at least one feature amount having a valid value equal to or higher than a predetermined threshold value as the feature amount to be used for personal recognition.

7. The information processing apparatus according to claim 1, wherein said selection means select, with a predetermined probabilistic weight, a predetermined number of feature amounts from the plurality of kinds of feature amounts detected by said detection means.

8. The information processing apparatus according to claim 7, wherein the probabilistic weight is determined on the basis of a distribution of the feature amounts detected by said detection means.

9. The information processing apparatus according to claim 1, further comprising hold means for holding groups of feature amounts with respect to individuals,

wherein said identification means includes collation means for performing collation between the at least one feature amount selected by said selection means and the groups of feature amounts held by said hold means to determine to whom the feature amount belongs.

10. The information processing apparatus according to claim 9, further comprising registration means for registering the at least one feature amount selected by said selection means in said hold means if the at least one feature amount selected by said selection means does not coincide with any of the personal feature amount groups held by said hold means.

11. The information processing apparatus according to claim 9, wherein if the feature amount selected by said selection means does not coincide with any of the personal feature amount groups held by said hold means, said identification means identifies the person based on some of the feature amounts different from the selected feature amount.

12. An information processing method, the method comprising:

performing by a processor the following steps;

a plurality of sensing steps of sensing a plurality of types of data from a space in which a person exists;

a detection step of detecting a plurality of kinds of feature amounts from the plurality of types of data sensed in the plurality of sensing steps;

a calculation step of calculating respective valid values of the plurality of kinds of feature amounts detected in the detection step;

a selection step of selecting at least one feature amount which has a relatively high valid value from among the plurality of kinds of feature amounts, on the basis of the respective valid values calculated in the calculation step; and

an identification step of identifying the person by using at least the selected one feature amount with the relatively high valid value,

wherein said calculation step includes;

obtaining normalized output level Vi of modules for detecting the feature amounts with respect to all classes i in a final layer of a hierarchical neural network;

calculating a sum total S=V₁+V₂+, . . . V_N of output levels in all the classes; and

calculating a valid value Yi of the feature amount in class i by computing Yi=ki×

Vi/S where ki represents the appearance frequency in class i.

13. A non-transitory computer-readable storage medium storing a computer-executable program for making a computer execute the information processing method according to claim 12.