Statistical Significance as an Aid to System Performance Evaluation

Abstract

Using forensic fingerprint identification as a testbed, a statistical framework for analyzing system performance is presented. Each set of fingerprint features is represented by a collection of binary codes. The matching process is equated to measuring the Hamming distances between feature sets. After performing matching experiments on a small data base, the number of independent degrees of freedom intrinsic to the fingerprint population is estimated. Using this information, a set of independent Bernoulli trials is used to predict the success of the system with respect to a particular dataset.

Cite

Text

Tu and Hartley. "Statistical Significance as an Aid to System Performance Evaluation." European Conference on Computer Vision, 2000. doi:10.1007/3-540-45053-X_24

Markdown

[Tu and Hartley. "Statistical Significance as an Aid to System Performance Evaluation." European Conference on Computer Vision, 2000.](https://mlanthology.org/eccv/2000/tu2000eccv-statistical/) doi:10.1007/3-540-45053-X_24

BibTeX

@inproceedings{tu2000eccv-statistical,
  title     = {{Statistical Significance as an Aid to System Performance Evaluation}},
  author    = {Tu, Peter Henry and Hartley, Richard I.},
  booktitle = {European Conference on Computer Vision},
  year      = {2000},
  pages     = {366-378},
  doi       = {10.1007/3-540-45053-X_24},
  url       = {https://mlanthology.org/eccv/2000/tu2000eccv-statistical/}
}