Types of Algorithms Used for Interpretation

Now that we know what constitutes a fingerprint and how it can be imaged , we need to know which types of algorithms are used. The algorithms used to match and enroll fingerprints fall under the following categories:

• Minutia-based

• Pattern-based

• Hybrid algorithm

Minutia-Based Algorithm

Earlier in this chapter, the minutia of fingerprints were described. As such, vendors who choose to use minutia-based algorithms will need to provide the highest-quality image possible. This way, the most minutia will be preserved. In addition, the presentation of the fingerprint after templating need not be as exact. The minutia in the template are compared to the raw template and matches are made. Minutia-based templates are relatively smaller than pattern-based matching templates.

This type of algorithm would be good to use in a situation where template sizing is important. For example, doing a match on a card would be more efficient with a minutia-based algorithm. For minutia-based fingerprint algorithms, only a small part of the finger image is required for verification. As such, it is ideal to have as much minutia as possible in the finger imaging area. Since just a portion of the minutia is required for verification, it would be ideal to use this algorithm where space restrictions impact the use and deployment of biometrics. Thus, a good imager for a minutia-based algorithm would be one that takes a high-quality image and has a large enough capture window for the relative core of the fingers to be imaged and captured.

Pattern-Based Algorithm

Pattern-based matching algorithms use both the micro and macro features of a fingerprint. When the macro features are utilized, the size of the image required for authentication, when compared to the size of the image needed for minutia-only requirements, is larger. Since only the macro features need to be compared, these types of algorithms tend to be fast and have a larger template size . They also require more of the image area to be present during verification. A good imager for pattern-based matching algorithms is one that has a high-quality camera and a large enough scanning surface to capture the important macro details.

Hybrid Algorithm

As the name implies, a hybrid algorithm uses the best features from the minutia-based algorithms and those from pattern-based matching. This algorithm is a good all-purpose algorithm, giving a good tradeoff between the accuracy of the minutia algorithm and the speed of pattern-based recognition. This algorithm would require the resulting template to be slightly larger than a minutia template, and smaller than a pattern-based matching algorithm. A high-quality optical sensor is best for this type of algorithm. It would offer a large enough image area, with very good quality for the images. The hybrid algorithm takes longer to enroll because of the use of both minutia and pattern-based recognition. Once this has occurred, the matching is actually faster than the minutia-based algorithm.

Which Algorithm Is Best?

This question can be answered positively for any algorithm given the right environment. With the right conditions, the pattern-based algorithm and minutia-based algorithm perform equally well. More important are the generalized cases for which a finger biometric may be used.

To decide on an algorithm, the use and implementation of the biometric need to be examined. If

• Template size is not important

• The relative speed differences between minutia-based and pattern-based algorithms are negligible

• The application does not require high throughput

then a pattern-based algorithm would work best. Otherwise, if

• Template size is important

• The relative speed difference is sufficiently meaningful given the volume of transactions

• The application has a high throughput

then a minutia-based algorithm will work best. Otherwise, if

• Template size is not important

• Faster matching is required than enrollment

• The application has high throughput

then a hybrid algorithm will work best.