mmm

Research Area

• Cyberspace
• Large- Scale System Theory
• Data Fusion & Mining 
• Energy and Resource Systems 
• Control Theories 
• Optimization

Recognition systems of one kind or another - among them character recognition systems, speech recognition systems, handwriting recognition systems, target recognition systems and pattern recognition systems - have been around for a long time. But what we are beginning to see today are recognition systems that are capable of performing tasks that could not be done in the past. Among examples of such systems are: 1. Computer virus detection system (IBM US Patent 5,675,711). This system employs a neural network classifier which is trained to detect both known and new viruses.

2. Eyeprint identification in ATM cash machines. In this system developed by NCR, a camera captures a digital record of a user's iris and can verify identity within seconds from a central database.

3. Supermarket checkout scanner (US Patent 5,673,089) which uses scent sensors to identify fruits and vegetables.

4. Molecular breathanalyzer that can detect diseases such as lung cancer, stomach ulcer and hepatitis at much earlier stages than currently used in radiological and laboratory tests.

5. Password authentication using typing biometrics.

6. MailJail software (Omron Advanced System) filters out unwanted junk e-mail. MailJail is a fuzzy-logic-based rule-based system which is customizable and is capable of learning user preferences about junk e-mail. 7. Seizure prediction actuator system (Georgia Institute of Technology). This system can recognize onset of an epileptic fit and can act to prevent it.

The quantum jump in the capabilities of today's recognition systems reflect three converging developments: (a) major advances in sensor technology; (b) major advances in sensor data processing technology; and (c) the use of soft computing techniques to infer a conclusion from observed data. Insofar as sensor technology is concerned, the advances in question relate to both availability and affordability. More specifically, such sensors as scent sensors, GPS sensors, MEMS sensors and DNA sensors did not exist in the past. When they did exist, they were unaffordable in terms of cost, weight, size or reliability. Today, sensor technology, and especially MEMS technology, provide us with a wide variety of ways in which information about a process can be obtained and processed at high speed, low cost and high reliability.

The employment of soft computing - which is a consortium of fuzzy logic, neurocomputing, evolutionary computing and probabilistic computing - is a key factor in the enhanced capabilities of recognition systems. To illustrate, the computer virus recognition system employs neurocomputing and machine learning; the password authentication system uses fuzzy logic; the MailJail software is fuzzy logic based; the seizure prediction system uses a combination of neurocomputing, wavelet analysis and fuzzy logic. In the future, most advanced recognition systems are likely to employ a combination of methodologies -rather than a single methodology - drawn from soft computing. A basic issue which is central to recognition technology relates to ways in which sets and, more generally, fuzzy sets can be defined. The principal modes are: (a) by a listing of elements; (b) by a recognition algorithm; (c) by a generation algorithm; and (d) by exemplification. An important part of the recognition process involves methods of passing from one mode to another. Fuzzy logic plays an essential role in this process. In the context of recognition, fuzzy logic is closely linked to the methodology of computing with words (CW). In coming years, recognition technology is likely to play a pivotal role in the conception, design, construction and utilization of information/intelligent systems. After all, recognition is one of the most basic facets of human reasoning and human cognition.