System and method for pose-angle estimation

(54) SYSTEM AND METHOD FOR POSE-ANGLE ESTIMATION

(75) Inventor: Mohamed Nabil Moustafa, Bayonne, NJ (US)

(73) Assignee: Identix Incorporated, Minnetonka, MN (US)

( * ) Notice: Subject to any disclaimer, the term ol this patent is extended or adjusted under 35 U.S.C. 154(b) by 54 days.

(21) Appl. No.: 10/175,929

(22) Filed: Jun. 20, 2002

(65) Prior Publication Data

US 2003/0235332 Al Dec. 25, 2003

(51) Int. CI.7 G06K 9/00; G06K 9/62;

G06E 1/00

(52) U.S. CI 382/159; 382/117; 340/5.53;

706/20

(58) Field of Search 382/115, 117,

382/118, 156-160, 282, 298, 253, 216; 340/5.53, 5.8-5.83, 553; 706/15, 20, 25

(56) References Cited

U.S. PATENT DOCUMENTS

5,239,594 A * 8/1993 Yoda 382/158

5,642,431 A 6/1997 Poggio et al 382/118

5,850,470 A * 12/1998 Kung et al 382/157

6,128,397 A 10/2000 Baluja et al.

6,144,755 A 11/2000 Niyogi et al.

6,429,699 Bl * 8/2002 Shi et al 327/103

OTHER PUBLICATIONS

Kohonen et al., "Engineering Applications ol the Sell-Or-
ganizing Map," IEEE 1996, pp. 1358-1384.*
Amari, "Training Error, Generalization Error and Learning
Curves in Neural Learning," IEEE 1995, pp. 4-5.*
Rowley et al., "Rotation Invariant Neural Network-Based
Face Detection," IEEE 1998, pp. 1-7.*
Lawrence et al., "Convolution Neural Networks for Face
Recognition," IEEE 1996, pp. 217-222.*

Rowley et al., "Rotation Invariant Neural Network-Based Face Detection," Proceedings ol IEEE Conlerence on Computer Vision and Pattern Recognition, pp. 1-7 (Jun. 1998). Li et al., "Support Vector Regression and Classification Based Multi-view Face Detection and Recognition," Proceedings IEEE International Conference on Automatic Face and Gesture Recognition, Mar. 2000.

Yan et al., "Ensemble SVM Regression Based Multi-View Face Detection System," Microsoft Research Technical Report MSR-TR-2001-09, Jan. 18, 2001.

Shimizu et al., "Head Pose Determination from One Image Using a Generic Model," Proceedings IEEE International Conference on Automatic Face and Gesture Recognition, Apr. 1998.

Kohonen, "Sell-Organizing Maps," ISBN 3-540-62017-6, Springer-Verlag 2 ed., 1997 (Title page, table ol contents and index only).

* cited by examiner

Primary Examiner—Mehrdad Dastouri

(74) Attorney, Agent, or Firm—Beyer Weaver & Thomas,

LLP

A system and method are disclosed for determining the pose angle ol an object in an input image. In a prelerred embodiment, the present system comprises a pose estimator having a prototype projector, a regression estimator, and an angle calculator. The prototype projector is prelerably adapted to reduce the input image dimensionality for laster lurther processing by projecting the input pixels ol the image onto a Sell-Organizing Map (SOM) neural network. The regression estimator is prelerably implemented as a neural network and adapted to map the projections to a pattern unique to each pose. The angle calculator prelerably includes a curve fitter and an error analyzer. The curve fitter is prelerably adapted to estimate the pose angle from the mapping pattern. The error analyzer is prelerably adapted to produce a confidence signal representing the likelihood ol the input image being a lace at the calculated pose. The system also prelerably includes two network trainers responsible for synthesizing the neural networks.

49 Claims, 7 Drawing Sheets