Systems and methods for facial age identification

Patent metadata

Field	Value
Publication number	US-9626597-B2
Application number	US-201414265846-A
Country	US
Kind code	B2
Filing date	Apr 30, 2014
Priority date	May 9, 2013
Publication date	Apr 18, 2017
Grant date	Apr 18, 2017

Abstract

Systems and methods are provided for acquiring classification functions for facial age identification. For example, a plurality of facial images associated with different ages are acquired; the facial images are assigned into a plurality of facial image collections based on at least information associated with a plurality of first age groups; for a first age group, one or more first facial image collections associated with one or more second age groups older than the first age group are acquired as positive samples; one or more second facial image collections associated with one or more third age groups younger than the first age group are acquired as negative samples; and training is performed based on at least information associated with the first positive samples and the negative samples to determine one or more classification functions for the first age groups.

First claim

The invention claimed is:

1. A method for acquiring classification functions for facial age identification, the method comprising:
acquiring a plurality of facial images associated with different ages;
assigning the facial images of the plurality of facial images into a plurality of facial image collections based on at least information associated with a plurality of first age groups, wherein each first age group of the plurality of first age groups corresponds to a facial image collection of the plurality of facial image collections, and wherein each facial image collection of the plurality of facial image collections includes one or more of the facial images;
for each first age group of the plurality of first age groups,
acquiring one or more first facial image collections associated with one or more second age groups older than that first age group as positive samples; and
acquiring one or more second facial image collections associated with one or more third age groups younger than that first age group as negative samples;

performing training based on at least information associated with the positive samples and the negative samples to determine one or more classification functions for each of the first age groups;
determining an estimated age of an additional facial image based on at least information associated with the one or more classification functions.

2. The method of claim 1 , further comprising:
determining weights of the positive samples and the negative samples of each of the first age groups with respect to a current age group based on at least information associated with age differences between the current age group and the positive samples and the negative samples of that first age group; and
wherein the performing training based on at least information associated with the first facial image collections and the second facial image collections to determine one or more classification functions for each of the first age groups includes:
performing training based on at least information associated with the positive samples and the negative samples of that first age group and the weights of the positive samples and the negative samples of that first age group to determine the one or more classification functions.

3. The method of claim 1 , wherein the acquiring the plurality of facial images associated with different ages includes:
acquiring a plurality of original images including at least human faces; and
acquiring the plurality of facial images based on at least information associated with the original images.

4. The method of claim 3 , wherein the acquiring the plurality of facial images based on at least information associated with the original images includes:
for each original image of the plurality of original images,
performing facial detection and eye positioning in that original image to acquire a facial position and an eye position; and
calibrating that original image based on at least information associated with the facial position and the eye position to acquire the plurality of facial images.

5. The method of claim 1 , wherein the performing training based on at least information associated with the positive samples and the negative samples to determine one or more classification functions for each of the first age groups includes:
extracting textural features of the positive samples and the negative samples of that first age group; and
performing training based on at least information associated with the textural features of the positive samples and the negative samples of that first age group and respective weights of the positive samples and the negative samples of that first age group to determine the classification functions.

6. The method of claim 5 , wherein the extracting textural features of the positive samples and the negative samples of that first age group includes:
extracting initial textural features of the positive samples and the negative samples of that first age group; and
performing Principal component analysis (PCA) dimensionality reduction on the initial textural features to acquire the textural features of the positive samples and the negative samples of that first age group.

7. The method of claim 1 , wherein the determining the estimated age group for the facial image based on at least information associated with the plurality of classification results includes:
calculating the estimated age group for the facial image based on at least information associated with the plurality of classification results and a span of that age group.

8. A method for facial age identification, the method comprising:
acquiring a facial image for facial age identification;
extracting textural features of the facial image;
classifying the facial image based on at least information associated with the textural features and one or more classification functions for one or more age groups;
acquiring a plurality of classification results, each classification result corresponding to an age group of the one or more age groups; and
determining an estimated age group for the facial image based on at least information associated with the plurality of classification results;
wherein:
a classification function of the one or more classification functions for a first age group of the one or more age groups is determined based on at least information associated with training of positive samples and negative samples of the first age group, the positive samples including one or more second facial image collections associated with a second age group older than the first age group, the negative samples including one or more third facial image collections associated with a third age group older than the first age group;

wherein weights of the positive samples and weights of the negative samples are determined for each of the age groups with respect to a current age group based on at least information associated with age differences between the current age group and the positive samples and the negative samples of that respective age group; and
wherein the classification function is determined by performing training based on at least information associated with the positive samples and the negative samples of that respective age group and the weights of the positive samples and the negative samples of that respective age group.

9. A system for acquiring classification functions for facial age identification, the system comprising:
one or more data processors; and
a computer-readable storage medium encoded with instructions for commanding the one or more data processors to execute operations including:
acquiring a plurality of facial images associated with different ages;
assigning the facial images of the plurality of facial images into a plurality of facial image collections based on at least information associated with a plurality of first age groups, wherein each first age group of the plurality of first age groups corresponds to a facial image collection of the plurality of facial image collections, and wherein each facial image collection of the plurality of facial image collections includes one or more of the facial images;
for each first age group of the plurality of first age groups,
acquiring one or more first facial image collections associated with one or more second age groups older than that first age group as positive samples; and
acquiring one or more second facial image collections associated with one or more third age groups younger than that first age group as negative samples;

performing training based on at least information associated with the positive samples and the negative samples to determine one or more classification functions for each of the first ag

Assignees

• Tencent Tech Shenzhen Co Ltd

Inventors

• Xiong Pengfei

Classifications

• G06V10/772

  Determining representative reference patterns, e.g. averaging or distorting patterns; Generating dictionaries · CPC title

• G06F18/28

  Determining representative reference patterns, e.g. by averaging or distorting; Generating dictionaries · CPC title

• G06V40/172 Primary

  Classification, e.g. identification · CPC title

• G06K9/6255 Primary

  Physics · mapped topic

• G06K2009/00322

  Physics · mapped topic