5. Conclusion
Old matching methods and even those are still being used, have
time-consuming procedures with no respect of hardware limitations.
However, the methods based on fingerprint ridge direction
pattern do not suffer from this problem. Another significant benefit
of the proposedmethod is the use of ergodic topology for HMM. This
approach overcomes the deficiencies of the former HMM matching
algorithm.
In general, the preferences and contributions of our proposed
HMM matching method to the previous one should be considered
as follows:
1. It is as fast as the previous HMM fingerprint matching method
with improved matching results.
2. The reference area of the suggested matching method is almost
of the same size as the previous one. However, it supplies much
information for the utilized HMM.
3. In the proposed method, all super states contain reference point
block.
4. There is more symmetry around reference point which leads to
more accuracy in similarity or likelihood computation.
5. There is no more need of extending 1-D HMM structure to a
pseudo 2-D one.
6. The proposed model is not a memoryless process and thus, it is
a generating case for a Markov model.
7. The intrinsic symmetrical structure of the proposed method
shows off in hidden state transition matrix.
8. It provides both HMM estimation and training without the zero
state transition problem