System and method for preventing enrollment of confusable patterns in a reference database

• US 5,754,977 A
• Filed: 03/06/1996
• Issued: 05/19/1998
• Est. Priority Date: 03/06/1996
• Status: Expired due to Term

• Current Assignee: Intervoice Limited Partnership
• Original Assignee: Intervoice Limited Partnership
• Inventors: Nguyen, Tuan K., Gardner, Marc A.
• Primary Examiner(s): Hudspeth, David R.
• Assistant Examiner(s): Storm, Donald L.
• Application Number:
  US08/611,709
• Time in Patent Office: 804 Days
• Field of Search: 395/2.53, 395/2.54, 395/2.52, 395/2.49, 395/2.47, 395/2.48, 395/2.45, 704/251, 704/235, 704/231, 704/200
• US Class Current: 704/243
• CPC Class Codes: G06F 18/28 Determining representative reference patterns, e.g. by averaging or distorting; Generating dictionaries
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 2015/0638 Interactive procedures

Abstract

A system and method for increasing the reliability of data matches between data stored in a database and newly provided data. The system relies on the prescreening of newly provided data patterns to insure that any newly provided data pattern is not ambiguous with existing data patterns. A threshold level is calculated for each newly provided data pattern and the invention determines which stored data pattern is closest to, or most likely to be confused with, the newly provided data pattern. The invention temporarily prohibits the new data pattern from being added to the database if it is likely to be confused with a stored data pattern. A user can then choose either to change the new data pattern to a different pattern or to allow the potentially confusing new data pattern to be added to the database. The calculated threshold levels can be set for both individual users and groups of users. The data patterns can consist of speech, images, graphs or other data.

Claims

1. The method of adding data patterns to a database, said database of patterns used for matching a subsequently inputted data pattern to a previously stored data pattern to obtain a data pattern which closely matches said subsequently inputted data pattern, said method comprising the steps of

accepting said subsequently inputted data pattern as a new pattern to be added to said database;

determining if any of said previously stored data patterns are within a certain match tolerance of said new pattern;

ordering said previously stored data patterns in a temporary list depending upon said new pattern; and

temporarily inhibiting the adding of said new pattern to said database under control of said determining step if said new pattern is within said certain match tolerance.

2. The method set forth in claim 1, wherein said determining step further includes the step of

determining the distance between said new pattern and a top data pattern on said ordered temporary list.

3. The method set forth in claim 1, wherein a user controls the input of said new pattern and wherein said temporarily inhibiting step includes the step of informing said user that said new pattern is within said certain match tolerance.

4. The method set forth in claim 1 wherein said previously stored data pattern is unique to a particular user.

5. The method set forth in claim 4 wherein said previously stored data pattern is common to a group of users.

6. The method set forth in claim 5 wherein said certain tolerance depends upon whether said stored data pattern is unique or common.

14. The system of claim 1 wherein said certain level is adjustable.

15. The invention set forth in claim 1 wherein said stored data patterns are unique to a particular user.

16. The invention set forth in claim 1 wherein said stored data patterns are common to a group of users.

17. The invention set forth in claim 1 wherein said calculating means includes:

means for comparing said accepted data pattern against data patterns unique to a particular user.

18. The invention set forth in claim 17 wherein said calculating means further includes:

means for comparing said accepted data pattern against data patterns common to a group of users.

19. The invention set forth in claim 18 wherein said threshold tolerance certain level is different between said comparison against said unique data patterns and said comparison against said common data patterns.

20. The system of claim 1 wherein said calculating means uses the algorithm:

##EQU2## where;

D=Distance between Current output and the Closest sample,

D_i =Coefficient delta,

C_s =Closest sample coefficient array

C_o =Current output coefficient array;

to determine if said newly provided data pattern is likely to be confused with said at least one stored data pattern.

21. The invention set forth in claim 1 wherein said data patterns are spoken words.

22. The invention set forth in claim 1 wherein said data patterns are images.

23. The invention set forth in claim 1 wherein said data patterns are graphs.

7. A system for extracting data from a database, said system comprising:

means for entering said database to extract a particular data file depending upon a match between a newly inputted data pattern and one of a plurality of data patterns previously stored in said database;

means for accepting a new data pattern for addition to said previously stored data patterns;

means for determining if any of said previously stored data patterns are within a certain match tolerance of said accepted new data pattern;

means for ordering said stored data patterns that are within said certain match tolerance in a temporary list; and

means controlled by said determining means for providing a signal when said new pattern is within said certain match tolerance.

8. The invention set forth in claim 7 wherein said certain match tolerance is different depending upon the particular previously stored data pattern.

9. The invention set forth in claim 7 wherein said signal providing means further includes:

means for temporarily inhibiting said adding of said new data pattern to said previously stored data patterns when said new pattern is within said certain match tolerance.

10. The invention set forth in claim 7, wherein said data patterns are spoken words.

11. The invention set forth in claim 7, wherein said data patterns are images.

12. The invention set forth in claim 7, wherein said data patterns are graphs.

13. A data recognition system comprising:

a plurality of stored data patterns, each pattern available for comparison to a subsequently provided data pattern for yielding a unique output depending upon a match between said subsequently provided data pattern and a previously stored data pattern;

means for accepting a data pattern for storage among said stored data patterns, said accepting means including;

means for calculating for each newly provided data pattern at least one stored data pattern which is most likely to be confused with said newly provided data pattern;

means for providing a list of said at least one stored data patterns ordered from the closest to the furthest from said newly provided data pattern;

means for determining a threshold tolerance between said newly provided data pattern and said stored data pattern determined to be most likely confused with said newly provided data pattern; and

means for at least temporarily delaying the storage of each newly provided data pattern when the threshold tolerance for that newly provided data pattern is below a certain level.