Improved Fuzzy Snakes Applied to Biometric Verification Problems

Abstract

Some types of biometric patterns can be represented as a collection of variable-length interconnected lines. This is the case of handwriting signature strokes, palmprint lines or infrared hand vein data. Typical variations in size, shape and orientation of these patterns for the same person make difficult to develop reliable biometric verification systems for them. Fuzzy snakes have been successfully applied to the off-line signature verification problem where the corresponding energy function is described by a set of fuzzy rules. In this paper, we extend the fuzzy shape-memory snake model by introducing a new external energy term: the difference between the angle of the tangent to the snake in a control point and the angle of the tangent to a specific stroke point (for all the strokes of the test pattern). Experimental results for both off-line signature and palmprint verifications have shown that the new fuzzy approach outperforms other snake models.