On the Difference between Optimal Combination Functions for Verification and Identification Systems

We have investigated different scenarios of combining pattern matchers. The combination problem can be viewed as a construction of a postprocessing classifier operating on the matching scores of the combined matchers. The optimal combination algorithm for verification systems corresponds to the likelihood ratio combination function. It can be implemented by the direct reconstruction of this function with genuine and impostor score density approximations. However, the optimal combination algorithm for identification systems is difficult to express analytically. We will show that this difficulty is caused by the dependencies between matching scores assigned to different classes by the same classifier. The experiments on the large sets of scores from handwritten word recognizers operating on postal images and biometric matchers (NIST biometric score set BSSR1) confirm the existence of such dependencies and that the optimal combination functions for verification and identification systems are different.

[1]  Adam Krzyżak,et al.  Methods of combining multiple classifiers and their applications to handwriting recognition , 1992, IEEE Trans. Syst. Man Cybern..

[2]  P. Jonathon Phillips,et al.  Models of large population recognition performance , 2004, Proceedings of the 2004 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2004. CVPR 2004..

[3]  Ching Y. Suen,et al.  A Method of Combining Multiple Experts for the Recognition of Unconstrained Handwritten Numerals , 1995, IEEE Trans. Pattern Anal. Mach. Intell..

[4]  Roland Auckenthaler,et al.  Score Normalization for Text-Independent Speaker Verification Systems , 2000, Digit. Signal Process..

[5]  Chaohong Wu,et al.  Iterative Methods for Searching Optimal Classifier Combination Function , 2007, 2007 First IEEE International Conference on Biometrics: Theory, Applications, and Systems.

[6]  Gyeonghwan Kim,et al.  Bankcheck Recognition Using Cross Validation Between Legal and Courtesy Amounts , 1997, Int. J. Pattern Recognit. Artif. Intell..

[7]  Mübeccel Demirekler,et al.  Undesirable effects of output normalization in multiple classifier systems , 2003, Pattern Recognit. Lett..

[8]  Robert P. W. Duin,et al.  Classifier Conditional Posterior Probabilities , 1998, SSPR/SPR.

[9]  Anil K. Jain,et al.  Large-scale evaluation of multimodal biometric authentication using state-of-the-art systems , 2005, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[10]  UedaNaonori.  Optimal Linear Combination of Neural Networks for Improving Classification Performance , 2000 .

[11]  Jiri Matas,et al.  On Combining Classifiers , 1998, IEEE Trans. Pattern Anal. Mach. Intell..

[12]  Sung Bum Pan,et al.  Fusion for Multimodal Biometric Identification , 2005, AVBPA.

[13]  Shaogang Gong,et al.  Audio- and Video-based Biometric Person Authentication , 1997, Lecture Notes in Computer Science.

[14]  Venu Govindaraju,et al.  Utilizing Independence of Multimodal Biometric Matchers , 2006, MRCS.

[15]  Naonori Ueda,et al.  Optimal Linear Combination of Neural Networks for Improving Classification Performance , 2000, IEEE Trans. Pattern Anal. Mach. Intell..

[16]  B. V. K. Vijaya Kumar,et al.  Unified decision combination framework , 1998, Pattern Recognit..

[17]  Venu Govindaraju,et al.  Classifier Combination Types for Biometric Applications , 2006, 2006 Conference on Computer Vision and Pattern Recognition Workshop (CVPRW'06).

[18]  Anil K. Jain,et al.  Statistical Pattern Recognition: A Review , 2000, IEEE Trans. Pattern Anal. Mach. Intell..

[19]  Djamel Bouchaffra,et al.  A Methodology for Mapping Scores to Probabilities , 1999, IEEE Trans. Pattern Anal. Mach. Intell..

[20]  Gyeonghwan Kim,et al.  A Lexicon Driven Approach to Handwritten Word Recognition for Real-Time Applications , 1997, IEEE Trans. Pattern Anal. Mach. Intell..

[21]  B. Ripley,et al.  Pattern Recognition , 1968, Nature.

[22]  Sharath Pankanti,et al.  The relation between the ROC curve and the CMC , 2005, Fourth IEEE Workshop on Automatic Identification Advanced Technologies (AutoID'05).

[23]  Aaron E. Rosenberg,et al.  Speaker background models for connected digit password speaker verification , 1996, 1996 IEEE International Conference on Acoustics, Speech, and Signal Processing Conference Proceedings.

[24]  J.P. Campbell,et al.  Allowing good impostors to test , 1997, Conference Record of the Thirty-First Asilomar Conference on Signals, Systems and Computers (Cat. No.97CB36136).

[25]  Konstantinos Sirlantzis,et al.  On the relation between dependence and diversity in multiple classifier systems , 2005, International Conference on Information Technology: Coding and Computing (ITCC'05) - Volume II.