Support

Support Options

Submit a Support Ticket

Projects

Language Technologies

Multilingual OCR

(Arabic, Indic, and CJK scripts)

 

Our objective is to automatically recognize handwritten and mixed-type documents in multiple languages. Our current focus is on pre-processing of documents and development of novel features for handwriting recognition.

 

Historical Documents

(restoration, transcription, indexing)

 

We are exploring the development of computational methodologies and applications that will enable user-friendly digital access to literary and historical document images in Sanskrit. Since OCR technology for handwritten Indic documents has not achieved a level of accuracy to be of practical use, we take advantage of the availability of machine readable transcription of these historical texts and develop techniques to successfully correlate text with its image.

Tablet/PDA Interfaces

(online recognition, gestures)

 

In collaborative research settings, academic researchers rely on both public and personal information, supplementing the study of technical literature with personal conversations and messages with other researchers, as well as individual notes. Although handwriting is still essential for much research-related “thinking,” understanding and problem solving activities in science and engineering, handwriting analysis is still an open problem when dealing with unconstrained environments and large lexicons. Hence, records of research design work which can prove to be very valuable for teaching and for future research are getting lost. We are investigating a new kind of scientific information interface which involves the use of a touch screen device such as a tablet, supporting sketching and handwriting input to access, annotate, store and retrieve research material. Personal research materials can therefore be organized and shared selectively in a reliable social networking environment by academic researchers.
 

White Board Technologies

(video, distance learning)
 

Editorial Gesture Recognition

(collaborative document editing)

Medical Informatics

(forms, prescriptions, lab reports)

 

Emergency Medical Systems (EMS) agencies, as first responders in medical emergency scenarios, are uniquely positioned to provide valuable data for surveillance and early warning systems. The PCR is used to gather vital patient information that is used by health care administrators as a resource to identify trends through macro analysis. Currently, the PCR data exists mostly on paper forms and the process of keying this data into a database that can be processed and mined for information can take up to several years. Technology being developed in the lab is enabling real-time collation and analysis of the data.

Web Security

(CAPTCHAs)

 

We are developing new technologies to beat the spambots by exploiting the difficulty that computers have with recognizing joined-up handwriting. The hope is that switching from text-based verification systems to systems that use computer-generated handwriting will make many web services more secure.
 

Search Engines

(handwriting, multilingual)

Writer Accent (Applications: digital library, security)

 

We are exploring two hypotheses with respect to analysis of an individual’s handwriting: (i) analogous to speech, accent is a trait that is also present in handwriting and can potentially be used as a novel soft biometric, (ii) an individual’s handwriting is a mixture of writing style influences (right to left versus left to right, short strokes with rapid pen lifts, curves, etc.) and can be decomposed into the constituents.
 

Biometrics

 

Modalities

(Face, fingerprint, gait, iris, signature)

.

Cancelable Biometrics

(private templates, cryptography)

 

Our focus is on Integrating Privacy Preserving Biometric Templates and Efficient Indexing Methods. While people are usually willing to submit their biometric information to government agencies, they are less likely to do so for commercial companies without a guarantee of privacy protection. Hence the need for (1) privacy of the biometrics in the database such that it can withstand malicious attacks and (2) fast matching of a biometric against a large database of stored biometric readings. This project is developing "cryptographically hidden" biometric templates which will lend themselves to fast searches.

People Tracking

(smart environments)

 

The ability to recognize, reason and retrieve information about human activity is critical to many applications, such as ubiquitous health-care monitoring and assistive technologies. The idea is that people can go about their normal activities without being subject to a ’pause and declare’ routine or the burden of wearing RFID tags or badges. Instead, evidence of identity of individuals accrues, or evolves, over the course of events as they get detected on unobtrusive biometric devices. We envision a ‘smart space’ instrumented by some sensors fixed to the physical environment in key locations while others are mounted on mobile ground or aerial robots and deployed on an ad hoc basis. to precisely locate individuals of interest as and when needed
 

Gaze Tracking

(assisted living, marketing)

Facial Expression Analysis

(deceit detection, impaired driving)

 

We are creating a new composite biometric that includes the static face appearance features as well as the behavioral expression features that will, in remote authentication applications, verify the humanness, liveness, and the identity of a user in a single interaction. This innovation would improve the trust and reliability of face recognition based authentication systems where the user might be operating from an unmonitored and remote location. We are also examining behavioral and physiological variables such as eyes gaze shifting for hotspot detection in counter terrorism scenarios. We envision automated tools to locate hotspots in video sequences which may be highly informative needing detailed human interpretation.
 

Emotion Analysis

(fMRI)

Scene Understanding

(locating people in crowds)

 

We are developing a 3D model based automated system for analysis of vehicle crash scenes to be used by emergency first responders. Our objective is to automatically determining the condition of the people: their physical condition (heartbeat and breathing rates) and soft biometrics traits such as approximate age and gender, and possibly their identities.

Fusion (modalities, classifiers)

 

(Application: video surveillance) We are taking a fresh view of biometric fusion where multiple frames and multiple views, rather than simply the multiple modalities. Sequences of video frames on a person’s activity are often available. The individual frames are acquired by possibly different cameras or even a single camera capturing multiple poses. Fingerprint probes can be also available in multiple scans for redundancy and quality control. The data from the multiple frames and views can be fused for reliable authentication of individuals. This represents a transformational shift in biometric fusion research and calls for biometric matchers to output additional information (e.g., estimated camera angle, image quality) for use by fusion algorithms.