Back to results
US 7,436,985 B2Granted
Personal identity authentication process and system
AssigneeOmniperception Limited
InventorJosef Kittler
Life of this patent
Estimating the term…
Abstract
A personal identity authentication process and system use a class specific linear discriminant transformation to test authenticity of a probe face image. A ‘client acceptance’ approach, an ‘imposter rejection’ approach and a ‘fused’ approach are described.
Claims31
A personal identity authentication process comprising the steps of:
- using linear discriminant analysis (LDA) to derive a class-specific linear discriminant transformation ai from N vectors zj (j=1,2 . . . N) defining respective training images, there being m different client classes of said training images, each said client class containing training images of a single client, with the ith class ωi containing a respective number Ni of said training images such that
- N
- =
- ∑
- i
- =
- 1
-
- m
-
- N
- i
- ,
- projecting a vector zp defining a probe image onto said class-specific linear discriminant transformation ai,
- comparing the projected vector
- a
- i
- T
-
- z
- p
- with a reference vector for the ith client class ωi and
- evaluating authenticity of the probe image with respect to the ith client class in dependence on the comparison.
20 more elements
A personal identity authentication process comprising the steps of:
- providing a class-specific linear discriminant transformation ai derived by applying linear discriminant analysis (LDA) to N vectors zj (j=1,2 . . . N) defining respective training images, there being m different client classes of said training images,each said client class containing training imager or a single client with the ith client class ωi containing a respective number Ni of said training images such that
- N
- =
- ∑
- i
- =
- 1
- m
-
-
- N
- i
- ,
- projecting a vector zp defining a probe image onto said class-specific linear discriminant transformation ai,
- comparing the projected
- a
- i
- T
-
- z
- p
- with a reference vector for the ith client class ωi and
- evaluating authenticity of the probe image with respect to the ith client class in dependence on the comparison.
20 more elements
Dependencies are inferred from the claim text.