Research Experiences for Undergraduates in

Information Processing and Decision Making for Intelligent and Secure Environments

Open House 2007

Twelve undergraduate participants of the REU summer program participated in the REU Open House 2007 to present their research projects.

• Passive Inhabitant Tracking and Behavior Anomaly Detection

  Two REU students were involved in the development of technologies that allow the non-intrusive tracking of inhabitants in a home and the modeling of their normal behavior patterns for the purpose of detecting unusual behavior. The motivation behind this is that the detection of unusual behavior can address some of the most pressing challenges in security sensitive as well as in home environments for the elderly. In particular, in the former anomalous behavior would indicate a potential threat while in the latter it can be used to automatically detect changes due to injury or mental changes and thus be used to provide early warning signs and make living in the home safer for the elderly.

  In this project, the two students worked collaboratively on developing a framework for behavior modeling and anomaly detection using Hidden Markov Models and Active LeZi algorithms. In addition, each student also implemented basic sensing techniques for a different sensor system, in particular for an omnidirectional vision system and a system consisting of an array of passive infrared sensors. The figure shows on the left the view of the omnidirectional camera system used for tracking and on the right a schematic of the infrared sensor layout and the embedded tracking and behavior model for the initial area. The goal of this project was to develop a robust system for voice control for intelligent environments. Such a voice control interface addresses one of the important issues in smart homes, namely the availability of an easy to use and natural interface between the human inhabitant and the devices in the home. The main challenges that had to be addressed for this project were the fact that the system has to operate in the context of ambient noise and in an environment where multiple people might have conversations or where radio or television might introduce conversations that can be on arbitrary topics. The latter leads to the problem that any set of key words used to identify commands might occur naturally during conversation but should be interpreted as commands only when they are intended as such.

• Intelligent Air Conditioning and Heating Control

  Another issue of increasing importance in home and commercial environments is energy efficiency. Since a large amount of energy is used for heating and air conditioning, the control of these devices is of major importance. In this project one REU student investigated a framework for the automatic optimization of heating and air conditioning controls. The system uses a predictive neural model to regulate the settings of the heating and air conditioning system such as to maintain the optimal level according to the comfort model of the persons occupying the space and is built to accommodate a component that will automatically learn the coefficients of the comfort model. The basic model is shown in the figure.

  In addition to designing the controller model, the student also designed and implemented a simulation environment that accurately simulated heating and air conditioning in a flexible configurable home environment, and that will be used to investigate the effectiveness of the intelligent heating control architecture.

• Intelligent Inventory Control for a Smart Kitchen

  This project built on previously developed hardware and software and was aimed at developing a system that can automatically track items and their amounts as they are put in and removed from a weight sensor and RFID enabled shelf system in a smart kitchen. One REU student worked on this project and developed the software architecture and interfaces to facilitate the item and quantity tracking as well as an interface between this architecture and a recipe system that allows to retrieve recipes that can be completed with the items available in the system. In addition, the student implemented algorithms to determine the sector in which items were placed on the shelf using the noisy sensor information and a graphical user interface that combined the tracking and the recipe recommendation system in a user friendly manner.

  The figure shows a schematic of the load assessment used in the system.

• Vein Pattern Recognition Technologies for Secure Home Entry

  Secure and intelligent environments increasingly rely on secure authentication technologies and in particular on biometrics. In this project one REU participant investigated different biometric authentication technologies and then focused on the development and construction of a vein pattern recognition and authentication system. Vein patterns are unique across individuals, change relatively little over an adult's lifetime and are very difficult to falsify, making them a very effective for biometric authentication. As opposed to techniques like fingerprint identification, however, vein pattern based systems are still relatively rare and very expensive. As part of this project the student constructed hardware and software for a relatively low cost system consisting of a near infrared camera, a set of LEDs to provide illumination in the frequency band most appropriate for the detection of the hemoglobin in veinous blood, and software to acquire the images and extract basic vein patterns.

The figure shows schematics of the basic hardware setup as well as a number of pictures taken by the implemented system, illustrating its capability to extract basic vein pattern information.

• Intelligent Service Robot Control

  Another important branch of technologies for intelligent environments is the one of service robots and in particular of intelligent control schemes that can deal with the uncertainty inherent in home environments. In this project one REU student investigated robot localization techniques and developed a scheme for the automatic mapping of the environment using a simple Roomba cleaning robot and a set of virtual walls to establish anchor points that are detectable using the robot's relatively simple sensors. The goal here was to deal with the uncertainty using a low cost platform and the resulting system used carefully designed navigation patterns to incrementally build a library of known anchor points in the environment that the robot uses to correct its location estimate.

  The figure shows the basic robot interface to drive the Roomba robot and an example navigation pattern for the construction and use of the anchors.

• Adaptive and Secure Personal Computer Interfaces

  Personalization is an important capability for computer systems in the context of providing the best access to information possible and allowing computer systems to be used efficiently by different users. The importance of this is further increased by the move towards pervasive computing systems and the need for privacy and security when dealing with personal information. Throughout the summer, two REU participants investigated different aspects of personalized computer interfaces with one of them focusing on the development of a portable personal desktop and application setup that allows a user to bring their own setup to a different computer without leaving any information after logging out. The other project investigated technologies for the automatic adaptation of the computer interface using machine learning to address the requirements of the particular user with particular aim at users with disabilities.

  The figure shows the basic operational flow of Zapplication, a portable personal computer desktop, and a sample desktop for the prototype system.

• Advanced Video Compression

  One of the most information and data intensive modalities in secure environments is surveillance and communications video, making it essential to have high performance and low cost video compression algorithms. This requirement is further moved to the forefront in pervasive and mobile settings where bandwidth, energy, and computing power is at a premium. In this project, one REU participant investigated modern video compression technologies and developed an enhancement to a preexisting video encoder.

• Interactive Multi-Point Video Communications for Law Enforcement

  In this project, one REU participant investigated interactive, mobile, pervasive video and audio technologies in the context of COPSE, a system for information exchange for law enforcement. The overall system is intended to provide reliable, real time video, audio, and data communications between police officers, their vehicles, and the police station. Building on a preexisting prototype, the student developed components that facilitate bidirectional video and audio communications between hand-held devices and a more powerful module intended for use on police cars. The main challenges here lay in developing components that provide sufficient quality yet have low computational overhead and that are reliable in the context relatively low power and low bandwidth wireless, battery operated devices in environments with significant wireless interference.

• Next Generation Multi-View Videonteractive Multi-Point Video

  Another issue arising in the context of visual surveillance is the integration of visual information into easier to analyze representation for next generation video systems. One of the issues arising in this is the integration of video from multiple cameras overlooking the same scene into a single representation that can be manipulated to obtain the best information. To address this, two REU participants investigated aspects of multi-view video and developed a video coding scheme and a viewer that allow the efficient transmission of multiple video streams taken from different viewpoints and their integration into a display in which the user can manipulate the viewpoint used for the display during the replay of the video.

  The figure shows the developed interface displaying video containing camera feeds from three different orientations.

For additional information regarding this program please contact:

Manfred Huber
University of Texas at Arlington
Department of Computer Science and Engineering
Box 19015
Arlington, TX 76019-0015
Email: reu@cse.uta.edu

* This project is funded by NSF grant CNS-0649229