From a theoretical point of view, Hidden Markov Models (HMMs) and Dynamic Bayesian Networks (DBNs) are similar, still in practice they pose different challenges and perform in a different manner. In this study we present a comparative analysis of the two spatial-temporal classification methods: HMMs and DBNs applied to the Facial Action Units (AUs) recognition problem. The Facial Action Coding System (FACS) developed by Ekman and Friesen decomposes the face into 46 AUs, each AU being related to the contraction of one or more specific facial muscles. FACS proved its applicability to facial behavior modeling, enabling the recognition of an extensive palette of facial expressions. Even though a lot has been published on this theme, it is still difficult to draw a conclusion regarding the best methodology to follow, as there is no common basis for comparison and sometimes no argument is given why a certain classification method was chosen. Therefore, our main contributions reside in discussing and comparing the relative performance of the two proposed classifiers (HMMs vs. DBNs) and also of different Region of Interest (ROI) selections proposed by us and different optical flow estimation methods. We can consider our automatic system towards AUs classification an important step in the facial expression recognition process, given that even one emotion can be expressed in different ways, fact that suggests the complexity of the analyzed problem. The experiments were performed on the Cohn-Kanade database and showed that under the same conditions regarding initialization, labeling, and sampling, both classification methods produced similar results, achieving the same recognition rate of 89% for the classification of facial AUs. Still, by enabling non-fixed sampling and using HTK, HMMs rendered a better performance of 93% suggesting that they are better suited for the special task of AUs recognition.
The use of sensor networks has been proposed for military surveillance and environmental monitoring applications. Those systems are composed of a heterogeneous set of sensors to observe the environment. In centralised systems the observed data will be conveyed to the control room to process the data. Human operators are supposed to give a semantic interpretation of the observed data. They are searching for suspicious or unwanted behaviour. The increase of surveillance sensors in the military domain requires a huge amount of human operators which is far beyond available resources. Automated systems are needed to give a context sensitive semantic interpretation of the observed kinematic data. As a proof of concept two automatic surveillance projects will be discussed in this paper. The first project is about a centralised system based on the AISAutomated Identification System which will be used to monitor ship movements automatically. The second project is about a decentralised system composed of a network of cameras installed at a military area. There is a need for a surveillance system along the coast of Europe. There is an increase of illegal drugs transport from the open sea, intrusion of boat refuges, illegal fishing, pollution of the sea by illegal chemical and oil pollution by ships. An automated sensor system is needed to detect illegal intruders and suspicious ship movements. Vessels fitted with AIS transceivers and transponders can be tracked by AIS base stations located along coast lines or, when out of range of terrestrial networks, through a growing number of satellites that are fitted with special AIS receivers. AIS data include a unique identifier of a vessel and kinematic data such as its position, course and speed. The proposed system enables identification, and tracking of vessels and to detect unwanted or illegal behaviour of ship movements. If ships violate traffic rules, enter forbidden areas or approach a critical infrastructure an alert will be generated automatically in the control room. Human operators start an emergency procedure. The second project is about a network of cameras installed at a military area. The area is monitored by multiple cameras with non-overlapping field of views monitored by human operators. We developed an automated surveillance system. At the entrance gate the identity of visitors will be checked by a face recognition system. In case of intruders, unwanted behaviour, trouble makers the emotional state of the visitor will be assessed by an analysis of facial expressions using the Active Appearance model. If unwanted behaviour is detected an alert is send the control room. Also license place of cars will be recognized using a system based on Neocognitron Neural Networks. Moving objects as persons and vehicles will be detected, localized and tracked. Kinematic parameters are extracted and a semantic interpretation of their behaviour is automatically generated using a rule based system and Bayesian networks. Cars violating the traffic rules or passing speed limits or entering forbidden areas or stopping/parking at forbidden places will be detected. A prototype of a system has been developed which is able to monitor the area 24 hours a day, 7 days a week.