6. Discussions & conclusions
In this work we presented a novel and robust identification approach
based on a DHMMK fromstatic lip information. Themain results
are:
• The performance of the DHMMK is significantly superior to that of
DHMM or SVM for all three datasets, proving that the use of static
lip information is a good biometric modality for the identification.
• The DHMMK performance improvement is greater for the more
challenging PIE and RaFD datasets, indicating a greater robustness to