Person identification based on movement information

• US 6,101,264 A
• Filed: 11/26/1996
• Issued: 08/08/2000
• Est. Priority Date: 03/15/1994
• Status: Expired due to Term

• Current Assignee: Hanger Solutions, LLC (IP Investments Group LLC)
• Original Assignee: Fraunhofer Gesellschaft Zur Foerderung Der Angewandten Forsching E.V.
• Inventors: Wagner, Thomas, Boebel, Friedrich G., Bauer, Norbert
• Primary Examiner(s): Couso, Jose L.
• Assistant Examiner(s): Dang, Duy M.
• Application Number:
  US08/716,247
• Time in Patent Office: 1,351 Days
• Field of Search: 382/112, 382/118, 382/115, 382/279, 382/282, 382/321, 395/20-25
• US Class Current: 382/115
• CPC Class Codes: G06V 40/20 Movements or behaviour, e.g. gesture recognition recognition of facial expressions G06V40/16
• CPC Class Codes: G06V 40/70 Multimodal biometrics, e.g. combining information from different biometric modalities
• CPC Class Codes: G07C 9/37 using biometric data, e.g. fingerprints, iris scans or voice recognition

Abstract

A method of identification of an individual is based on movement information. Movement information from an image sequence that results when a typical or predetermined movement is carried out by the individual to be identified is recorded and evaluated by means of optical flow in order to ensure high recognition dependability. Complementary audio and static image information may also be evaluated.

Claims

1. A process for a positive identification of an individual from a set of identifiable individuals, comprising the steps of:

(i) generating a set of classes at least corresponding to the set of identifiable individuals, each class having features which correspond to characteristics of an individual in said set of identifiable individuals;

(ii) providing a camera for recording an image sequence as movement information;

(iii) recording said image sequence with said camera, said sequence originating from at least one of a typical and predetermined movement action associated with a so far unidentified individual;

(iv) evaluating said recorded image sequence through an optical flow thereby producing a feature vector representing movement information contained in the recorded image sequence;

(v) comparing said produced feature vector with at least one class of said set of classes;

(vi) generating a positive identification signal corresponding to a certain class of said set of classes having features corresponding to said feature vector representing said movement information.

2. Process according to claim 1, wherein said predetermined movement action is a movement of lineaments of the individual to be identified when speaking a predetermined word or sentence.

3. Process according to claim 1, wherein said typical movement action is at least one of a walk and a head movement of the individual to be identified.

4. Process according to claim 1, wherein the typical movement action is recorded in combination with a predetermined movement action.

5. Process according to claim 1, wherein the recording of said at least one of a typical and predetermined movement action is combined with a recording of static images, said recorded static images are evaluated in addition to said evaluating of said recorded image sequence.

6. Process according to claim 5, wherein the recording of said static images is combined with a recording of audio information, said recorded audio information is evaluated in addition to said evaluating of said recorded images.

7. Process according to claim 1, wherein the evaluation of said image sequence is combined with an evaluation of audio information.

8. Process according to claim 1, wherein said set of classes is provided for comparing with an unknown feature vector corresponding to the so far unidentified individual, said unknown feature vector being evaluated through said optical flow, wherein one adjoint prototype obtained by a learning process is calculated for each class for comparing with said unknown feature vector.

9. A learning process for a process for a positive identification of an individual from a set of identifiable individuals, comprising the steps of:

(i) generating at least one class in a set of classes at least corresponding to the set of identifiable individuals, each class having features which correspond to characteristics of an individual in said set of identifiable individuals;

(ii) providing a camera for recording an image sequence as movement information;

(iii) recording said image sequence with said camera, said sequence originating from at least one of a typical and predetermined movement action associated with an individual to be represented by said at least one class;

(iv) evaluating said recorded image sequence through an optical flow thereby producing a feature vector representing movement information contained in the recorded image sequence; and

calculating an adjoint of said feature vector;

(v) providing said produced adjoint as said at least one class of said set of classes.

10. Learning process for an identification process according to claim 9, wherein a spectrogram is determined from an acoustic signal, said spectrogram is translated into a feature vector and converted into an adjoint.

11. Learning process according to claim 9, wherein said feature vector is obtained by interposing a Fourier transform.