The Centre for Pattern Analysis and Machine Intelligence (CPAMI) will conduct multidisciplinary research and development activities in six research areas related to intelligent systems and data analysis as illustrated in the following image.
CPAMI intends to break down barriers between the different disciplines and sciences, contributing to the field of intelligent systems and data analysis.
CPAMIs findings will have the potential to address numerous real world problems that traditional systems cannot resolve. The envisaged application directions include, but are not limited to, public security, manufacturing, transportation, assistive environments for seniors and people with disabilities, communication, eLearning, finance and web services. The following subsections describe in details the research areas of the centre:
While we would like our machine to understand and be aware of its environment, in actuality, a modern machine or a robot is limited by the sensors we give it and the software we write for it. Sensing is not perceiving. Sensors are merely transducers that convert some physical phenomena into electrical signals that the microprocessor can read. Perception is much more than parameter estimation; it involves the interpretation of complex data. CPAMI will carry out research and development activities in order to build machines capable of perceiving their surroundings. CPAMI will encompass two main focus groups to tackle the problems related to machine perception. These focus groups are Sensor Technologies focus group and the Machine Vision focus group. Both groups cooperate to produce smart sensors that understand and analyze the nature of the monitored environment and act accordingly.
The Sensor Technologies focus group's main task is to explore various forms of sensor technology for applications that advance science and improve energy efficiency. In addition, the group focuses on the technology that can survive in outdoor and harsh environments. Moreover, mobile sensor-related issues are one of the interests of this group.
The main task of the Machine Vision focus group is to push forward the boundaries of artificial vision research. Such efforts include investigating both active as well as passive vision methodologies. Active vision depends mainly on using laser scanners to extract 3D information for a given environment while passive vision techniques acquire digital images through which 3D information can be extracted. In the latter case, cameras are used to acquire images along with sensors to determine parameters. This particular point should be a collaborative area between both focus groups teams. That is because information inferred accurately by sensors can help researchers develop good vision systems. On the other hand, vision techniques can help researchers enhance the accuracy induced by sensors. The Machine Vision focus group team should work not only on projects that are related to mobile robots but on standalone vision systems like indoor and outdoor video surveillance systems.
Robotics and autonomous systems
Robotics and autonomous systems will be researched by four focus groups, namely, the Industrial Automation focus group, the Field Robotics focus group, the Service Robotics focus group, and the Cognitive Robotics focus group. These focus groups are small groups of researchers who have common research interests. Each group exchanges ideas, resources, and often meets periodically to discuss research progress.
The Industrial Automation focus group aims at advancing the state-of-the-art intelligent automated systems and industrial automation through the development of reliable solutions for real world problems.
The objective of the Field Robotics focus group is to conduct research towards the development and improvement of new generation field robots, and maintain cooperation with industry in projects of mutual interest and benefit. This group will focus on aerial and ground robotics, with an emphasis on 3D mapping & motion planning in unknown environments, nonlinear tracking and control, and multi-vehicle coordination.
The aim of the Service Robotics focus group is to explore the field of service robotics for private use focusing on the practical issues of personal assistive robots, cleaning robots, entertainment robots, etc. This focus group also intends to increase students’ awareness of current issues in service robotics and motivate them to work in teams with heterogeneous knowledge and skills.
The Cognitive Robotics focus group will investigate endowing robots with higher-level cognitive faculties, such as planning and natural language understanding. These faculties require a formal representation of the robot’s beliefs about the world and a reasoning mechanism that allows the robot to draw sound conclusions from what it knows. This research group will focus on mental models, robot learning and social interaction. The researchers involved in this focus group will study different topics related to the inner-nature of the cognitive robots such as computational situation awareness, knowledge representation, learning from communication and learning to communicate, reasoning and planning and social interaction. Endowing robots with these high level cognitive capabilities will to reason, act, and perceive in dynamic, partially known, and unpredictable environments in a robust manner. Moreover, these issues are highly relevant to the concerns of other research groups within the centre. Inter-group collaboration possibilities are evident and clearly recommended.
Cooperative intelligent systems
CPAMI will research and deploy computational intelligence mechanisms to mobile robotics and other autonomous intelligent control problems. CPAMI will work on providing the needed artificial intelligence to robots so that they can operate as individuals and groups in real world unstructured and open environments. CPAMI will research on the theoretical and practical aspects of constraint programming, Fuzzy Logic Systems, Neural Networks and Evolutionary Computation as well as their hybrids. Among the research directions of CPAMI is the construction of multiple classifier systems (MCS). A good understanding of how to build more sophisticated MCS and exploit various possibilities of extracting information from the environment will move us closer to achieving the original intent of machine intelligence, which is to automate the knowledge acquisition process.
CPAMI will also investigate different issues of sensor networks such as static sensor placement to maximize the coverage, energy-aware deployment of mobile sensors, routing, mobility, reliability, fault tolerance and adaptivity. The group will also conduct research on a variety of systems, networking and data management issues in data-centric sensor networks.
Human-machine interaction is the study of humans, machines, and the ways they influence each other. It is a multidisciplinary research area with contributions from the fields of human-computer interaction, artificial intelligence, robotics, natural language understanding, and social science (psychology, cognitive science, anthropology, and human factors). CPAMI will investigate verbal and non-verbal interaction techniques that can be used to provide novel and natural ways of interaction between humans and machines. This research will be conducted by three focus groups: Speech Understanding, Gesture Recognition and Emotion Classification.
The Speech Understanding focus group aims to develop various efficient algorithms for understanding the meaning of natural languages in different forms, spoken utterances, text, or multimedia and provides tools to utilize the meaning in different applications such as: machine translation, intelligent user interface, speech recognition and learning objects.
The Gesture Recognition focus group will address the gesture-based interaction as a process of real-time interaction, which is based on using an expressive movement of a part of the body, the hand or head in order to bring forward intentions and attitude.
The Emotion Classification focus group will investigate emotion expression through head/arm/eyelid movement focusing on how to distinguish one emotion from another. Emotions provide a rich and natural way of non-verbal (embodied) communication between human and machine.
Pattern recognition and image analysis
Research on pattern recognition and image analysis will include shape representation, texture analysis, visual inspection and biometrics.
Shape representation is a rapidly evolving discipline with roots in areas including image processing, computer vision, pattern recognition, artificial intelligence, and advanced mathematics. CPAMI will investigate how to create representations and descriptors, that are invariant to geometrical transformations, are needed to correctly match, recognize, or classify shapes.
Texture describes a wide variety of surface characteristics such as terrain, plants, minerals, fur and skin. Texture analysis is one of the fundamental aspects of human vision by which we discriminate between surfaces and objects. CPAMI will research on texture models, the main approaches to texture analysis and the applications of texture analysis.
Visual inspection is a common method of quality control, data acquisition, and data analysis. CPAMI will investigate real-world applications of visual inspection.
Biometrics refers to methods for uniquely recognizing humans based upon one or more intrinsic physical or behavioural traits. CPAMI will address both physiological features-based and behavioural features-based biometric techniques. Physiological features-based techniques include, but are not limited to, face recognition, thermogram, fingerprint, lip print, finger vein pattern, hand back vein, 3-D hand geometry, ear geometry, ear canal feedback/echo, iris scan, retinal scan, DNA, nail bed, nail radio-frequency identification (RFID), skin spectrum, dental biometrics, body odours, blood pulse/cardiac pulse, palm vein, palmprint and footprint recognition. Behavioural features-based techniques include, but not limited to, voice recognition, keystroke, signature, body movement, gait biometrics, gestures and tapping. A report will be produced to describe the reviewed approaches. Uni-modal and multi-modal techniques will be investigated.
Data mining and knowledge discovery
Data mining is the process of discovering prominent patterns from highly structured databases. Nowadays, the vast and overwhelming amount of information is considered as a challenge for the decision-makers. Since decision-makers are increasingly subjected to this information overload, selectivity and screening of information is seen as growing in importance. CPAMI aims at providing data mining-based techniques, which can dramatically shorten the time needed to extract useful knowledge from given data or information. These techniques include, but are not limited to, discover data anomalies, identify invalid and incomplete data, data classification and clustering, data association, meta-data extraction, information studies, information retrieval, information consolidation, performance evaluation, and planning & decision making.
CPAMI will also address knowledge systems focusing on the logical and technical foundation of knowledge discovery, such as:
- knowledge organization and acquisition
- knowledge discovery and pattern analysis for mixed-mode, relational and stochastic data
- cooperative knowledge discovery
- machine learning, machine reasoning and knowledge organization
- phrase based clustering, multi-clustering and aggregation
- hard, semi-fuzzy and fuzzy clustering methods and neural networks, and reinforcement learning
- multiagent learning and coordination