Biometric indexing and searching system
View Patent ↗A prototype biometric identification system is disclosed that indexes a biometric corpus into indexed-corpuses using a set of P prototypes before searching for a probe in a search corpus constructed based on the indexed-corpus. The system may index the biometric corpus based on the prototypes directly or based on prototype-typicality scores.
1. A biometric index and search apparatus that searches, based on a matching operation, for a probe in a biometric corpus stored in a database, comprising:
a plurality of prototype templates;
an index processor coupled to the database;
a search processor coupled to the index processor and the database; and
a typicality processor coupled to the index processor, the search processor, and the database, wherein:
the typicality processor performs the matching operation between templates in the biometric corpus and the prototype templates, and averages resulting match scores for each of the templates to generate a template-typicality score for each of the templates; and
the index processor generates a typicality-indexed corpus from the biometric corpus by dividing template-typicality scores of all the templates in the biometric corpus into typicality ranges, each typicality range corresponding to a typicality sub-corpus that has about a same number of templates as other typicality sub-corpuses, and each of templates in the typicality sub-corpus having a template-typicality score within a typicality range of the typicality sub-corpus.
2. The apparatus of claim 1 , wherein:
the typicality processor performs the matching operation between the probe and the prototype templates to generate probe-match scores, and averages the probe-match scores to generate a probe-typicality score; and
the search processor selects a typicality sub-corpus based on the probe-typicality score as the search corpus.
3. A biometric index and search apparatus that searches, based on a matching operation, for a probe in a biometric corpus stored in a database, comprising:
a plurality of prototype templates;
an index processor coupled to the database;
a search processor coupled to the index processor and the database; and
a typicality processor coupled to the index processor, the search processor, and the database, wherein:
the typicality processer performs the matching operation between each of the prototype templates and all other ones of the prototype templates to generate prototype-match scores, and averages the prototype-match scores for each of the prototype templates to generate a prototype-typicality score for each of the prototype templates; and
the index processor generates typicality proto-groups of the prototype templates by dividing prototype-typicality scores of all the prototype templates into P number of prototype-typicality ranges, each of the prototype-typicality ranges corresponding to one of the typicality proto-groups that has about a same number of prototype templates as other ones of the typicality proto-groups, prototype templates in each of the typicality proto-groups having a prototype-typicality score within a prototype-typicality range corresponding to each of the typicality proto-groups.
4. The apparatus of claim 3 , wherein the typicality processor:
performs the matching operation between all the templates in the biometric corpus and all prototype templates in each of the typicality proto-groups to generate template-match scores, and averages the template-match scores to generate template-typicality scores for each of the typicality proto-groups;
the index processor divides the biometric corpus into a set of proto-corpuses for each of the typicality proto-groups based on the template-typicality scores of each of the typicality proto-groups.
5. The apparatus of claim 4 , wherein the index processor divides the template-typicality scores of each of the typicality proto-groups into Q number of template-typicality ranges, each of the template-typicality ranges corresponding to one proto-corpus of the set of proto-corpuses, each of templates in the one proto-corpus having a template-typicality score within a template-typicality range corresponding to the one proto-corpus, each proto-corpus of the set of the proto-corpuses having about a same number of templates as other ones of the set of proto-corpuses.
6. The apparatus of claim 5 , wherein the search processor:
performs the matching operation between the probe and prototype templates of each of the typicality proto-groups to generate probe-match scores;
averages the probe-match scores to generate a probe-typicality score;
selects a proto-corpus from each set of proto-corpuses based on the probe-typicality score corresponding to the set of proto-corpuses; and
places templates that are in all selected proto-corpuses to form the search corpus.
7. A method for biometric indexing and searching based on a matching operation, for a probe in a biometric corpus comprising:
performing the matching operation between templates in the biometric corpus and prototype templates to obtain template-match scores;
averaging the template-match scores for each of a plurality of templates to generate a template-typicality score; and
dividing template-typicality scores of all the templates in the biometric corpus into typicality ranges, each typicality range corresponding to a typicality sub-corpus of a typicality-indexed corpus that has about a same number of templates as other typicality sub-corpuses corresponding to other typicality ranges, and each of templates in the typicality sub-corpus having a template-typicality score within a typicality range of the typicality sub-corpus.
8. The method of claim 7 further comprising:
performing the matching operation between the probe and the prototype templates to generate probe-match scores;
averaging the probe-match scores to generate a probe-typicality score; and
selecting a typicality sub-corpus based on the probe-typicality score as the search corpus.
9. A method for biometric indexing and searching based on a matching operation, for a probe in a biometric corpus comprising:
performing the matching operation between each of a plurality of prototype templates and all other ones of the prototype templates to generate prototype-match scores;
averaging the prototype-match scores for each of the prototype templates to generate a prototype-typicality score for each of the prototype templates; and
dividing prototype-typicality scores of all the prototype templates into P number of prototype-typicality ranges to generate typicality proto-groups, each of the prototype-typicality ranges corresponding to one of the typicality proto-groups that has about a same number of prototype templates as other ones of the typicality proto-groups, prototype templates in each of the typicality proto-groups having a prototype-typicality score within a prototype-typicality range corresponding to each of the typicality proto-groups.
10. The method of claim 9 , further comprising:
performing the matching operation between all the templates in the biometric corpus and all prototype templates in each of the typicality proto-groups to generate template-match scores, and averages the template-match scores to generate template-typicality scores for each of the typicality proto-groups;
dividing the biometric corpus into a set of proto-corpuses for each of the typicality proto-groups based on the template-typicality scores of each of the typicality proto-groups.
11. The method of claim 10 , further comprising:
dividing the template-typicality scores of each of the typicality proto-groups into Q number of template-typicality ranges, each of the template-typicality ranges corresponding to one proto-corpus of the set of proto-corpuses, each of templates in the one proto-corpus having a template-typicality score within a template-typicality range corresponding to the one proto-corpus, each proto-corpus of the set of the proto-corpuses having about a same number of templates as other ones of the set of proto-corpuses.
12. The method of claim 11 , further comprising:
performing the matching operation between the probe and prototype templates of each of the typicality proto-groups to generate probe-match scores;
averaging the probe-match scores to generate a probe-typicality score;
selecting a proto-corpus from each set of proto-corpuses based on the probe-typicality score corresponding to the set of proto-corpuses; and
placing templates that are in all selected proto-corpuses to form the search corpus.