Combining multiple independent sources of information for classification of devices under test

1. A method for determining that a unit is defective, said method comprising:

testing a unit with a plurality of defect detecting techniques;

generating weighted results of said defect detecting techniques, wherein said weighting is related to an accuracy of said techniques;

determining a probability that said unit has a defect, wherein said determination is made by combining said weighted results from said defect detecting techniques;

testing at least one standard unit having at least one known defect;

testing at least one standard unit which is substantially free of defects; and

wherein said accuracy of each said technique is determined by comparing results obtained by performing each said technique on each said at least one standard unit with anticipated results for each said at least one standard unit.

2. The method of claim 1 further comprising:

determining said accuracy of each said technique by comparing results for each said technique performed on a standard with expected results for said standard.

3. The method of claim 2 wherein said accuracy determination uses a Gaussian model.

4. The method of claim 1 further comprising:

calculating a likelihood for each said technique by mathematically combining said results for said technique and said accuracy for said technique.

5. The method of claim 4 further comprising:

calculating a probability that said unit has a defect by mathematically combining said likelihoods.

6. The method of claim 5 wherein said likelihoods are converted to probabilities and are combined using a Bayesian Voting Formula.

7. A system for evaluating a unit, said system comprising:

a plurality of testing techniques, wherein each said technique detects the presence of at least one characteristic, and wherein an accuracy of each said technique can be determined;

means for analyzing results of said techniques that determines a probability that said at least one characteristic is detected by using a combination of each technique's results and each said technique's accuracy, wherein said probability is used to evaluate said unit;

at least one standard having at least one known defect:

at least one standard which is substantially free of defects; and

wherein said accuracy of each said technique is determined by comparing results obtained by performing each said technique on each said at least one standard with anticipated results for each said at least one standard.

8. The system of claim 7 further comprising:

at least one standard having a known characteristic profile, wherein said at least one standard is used to determine said accuracy for each said technique.

9. The system of claim 7 wherein said at least one characteristic is a defect.

10. A computer readable storage medium storing computer executable code for evaluating a unit, said code comprising:

code for calculating a confidence factor for each of a plurality of tests;

code for combining results from each of said test with said confidence factor for that test producing a likelihood for each said test; and

code for combining said likelihoods to arrive at a determination regarding a unit tested by each said test.

11. The computer readable storage medium of claim 10 wherein said determination relates to whether said unit is defective.

12. The computer readable storage medium of claim 10 further comprising:

code for changing each said likelihood into a probability; and

code for combining said probabilities using a Bayesian Voting Formula.

13. The computer readable storage medium of claim 10 further comprising:

code for comparing results that each said test produces for a standard with results that are anticipated for said standard, wherein said comparison is used in said calculation of said confidence factors.

14. The computer readable storage medium of claim 10 wherein said determination produces at least one graph of a receiver operating characteristic.

15. A method of evaluating a unit, said method comprising:

providing a plurality of testing techniques for detecting defects in said unit;

testing at least one standard using each said technique, wherein said at least one standard has a known defect profile;

calculating a weighting factor for each said technique from results of said at least one standard testing; and

testing said unit using each said technique, wherein an acceptability of said unit is determined from a combination of said unit testing results, and said weighting factors.

16. The method of claim 15 further comprising:

producing a first feature set by testing a standard having at least one known defect;

producing a second feature set by testing a standard which is substantially free of defects; and

generating a statistical measurement of accuracy for each said technique from said first and second feature sets.

17. The method of claim 16 wherein said generating uses a Gaussian model.

18. The method of claim 15 wherein at least one of said techniques is an optical technique for the detection of defects on printed circuit boards.

19. The method of claim 15 wherein at least one said technique is an electrical conductivity technique.