Demographic analysis of facial landmarks
Inventors
Assignees
University of North Carolina at Wilmington
Publication Number
US-8705875-B1
Publication Date
2014-04-22
Expiration Date
2031-09-07
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Abstract
A facial image may be annotated with the plurality of facial landmarks. These facial landmarks may be points or regions of the face that are indicative, either alone or in combination with other facial landmarks, of at least one demographic characteristic. Demographic characteristics include, for example, age, race, and/or gender. Based on the demographic characteristic being analyzed, one or more of these facial landmarks may be selected and arranged into an input vector. Then, the input vector may be compared to one or more of the training vectors. An outcome of this comparison may involve in the given facial image being classified into a category germane to the analyzed demographic characteristic (e.g., an age range or age, a racial category, and/or a gender).
Core Innovation
The invention provides a method and system for analyzing facial images to estimate demographic characteristics such as age, race, and gender. Facial images are annotated with a plurality of facial landmarks, which are points or regions of the face that indicate at least one demographic characteristic. Depending on which characteristic is to be analyzed, particular facial landmarks are selected and arranged into an input vector for further classification.
The input vector, derived from the given facial image, is compared to a set of training vectors where each training vector is associated with known demographic characteristics. This comparison may use techniques such as support vector machines (SVM) for initial classification into categories (e.g., specific age ranges, races, or genders), and support vector regression (SVR) for further refinement within those categories (e.g., estimating an exact age within a predicted age range).
The problem being addressed is the lack of accuracy in existing approaches to determine demographic characteristics such as race, gender, or age from digital facial photographs. The invention improves accuracy by leveraging selected facial landmarks that are specifically chosen to represent anthropometric features relevant to the demographic characteristic being estimated, leading to better models and enhanced demographic classification and estimation.
Claims Coverage
The patent contains multiple independent claims that detail inventive methods and an article of manufacture for demographic analysis of facial landmarks. The coverage focuses on core processes for extracting, selecting, comparing, and classifying facial landmarks to estimate demographic information using computing devices.
Identifying and classifying facial landmarks for demographic estimation
A method that identifies an input vector of facial landmarks from a facial image, selects those landmarks based on anthropometric characteristics, performs a weighted comparison with training vectors mapped to categories such as age range, and classifies the given image accordingly. A second comparison is then performed within the classified category, for example, differentiating child age ranges using facial features indicative of bone expansion or hard-tissue elongation, and estimating the individual's age based on this second comparison.
Mapping training vectors to multidimensional space and regions for classification
A method where each training vector also maps to a region in m-dimensional space, one for each category, and the comparison process involves mapping the input vector to at least one region. Classification into a category is based on the region to which the input vector maps within this m-dimensional space.
Selecting and comparing application-specific subsets of facial landmarks
A method for determining which application to apply (such as age, race, or gender estimation) to a first input vector of facial landmarks. Based on this determination, a subset of landmarks is selected to form a second input vector. The second input vector is then compared to a set of training vectors, classification into a relevant category takes place, and further comparison within that category results in demographic estimation. For age estimation, features indicative of adult characteristics (wrinkling, hyper-pigmentation, thinning of fat padding) are used.
Non-transitory computer-readable medium for demographic analysis
An article of manufacture comprising a non-transitory computer-readable medium storing software instructions. When executed by a computing device, these instructions cause the device to perform operations including identifying facial landmarks, performing weighted comparisons to training vectors, classifying facial images into categories, and subsequently estimating age with attention to specific age-related features (e.g., bone expansion in children).
The inventive features covered include methods and systems for extraction, selection, comparison, and multidimensional classification of facial landmarks for demographic estimation, as well as software embodiments to implement such processes on computing devices.
Stated Advantages
Provides improved accuracy in estimating demographic characteristics such as age, race, and gender from digital facial photographs compared to prior methods.
Enables accurate age estimation with a mean absolute error of less than about 4 years, and in some testing scenarios, less than 3.96 years.
Allows reduction in the number of images to review in scenarios such as public security, by estimating demographics for filtering relevant photographs.
Documented Applications
Use by entities providing public security functions, such as law enforcement agencies, to narrow down digital photographs for identifying persons of interest based on estimated demographic characteristics.
Age verification in situations where legal status or qualification is dependent on age, such as verifying patrons' age in casinos and bars.
Verifying that participants in youth sports organizations are below or within a certain age range, as well as ensuring that professional athletes are above a certain age limit or within a specified range.
Interested in licensing this patent?