Speech recognition technology has made spoken interaction with machines feasible. Shrinking form factors point to speech as a crucial component of mobile computing and communication devices. However, no universal interface has yet been proposed for humans to effectively, efficiently and effortlessly communicate by speech with machines. This program will analyze human communication with a variety of devices and applications, and will design, implement and test a universal interface. This will consist of a metaphor (analogous to the desktop metaphor of GUIs), a set of universal interaction primitives (help, navigation, confirmation, correction etc.), and a graphical component for applications afforded a display. The team will conduct user studies to evaluate user acceptance and the transference of user skills across applications. The team will also create tools for rapid prototyping of compliant applications.

This program will study the integrated design, microfabrication and control of a prototype catalytic microreaction system for methanol fuel reforming. Such miniature fuel processors for in situ hydrogen production are of particular interest in compact fuel cells, which are currently being considered as alternative energy sources for high-end portable electronic devices such as cellular phones and laptops.

The following four main components will constitute the integrated miniaturized fuel processor:

• mixer/vaporizer;
• catalytic steam reformer with Cu/ZnO composite catalyst;
• water gas shift reactor assisted by a palladium membrane supported on a porous Cu/ZnO catalyst;
• integrated resistive heaters/sensors and control electronics.

These components will be microfabricated on a single silicon wafer thereby leading to an integrated "microplant-on-a-chip". The prototype will serve as a hydrogen source for compact miniature fuel cells which, in the near future, are expected to be the most viable power sources for portable electronic applications.

Due to their increasing complexity, the design and manufacture of high-quality, low-cost system-on-chip (SoC) ICs will require effective hierarchical test generation and fault simulation tools. Such tools require a fault model that can accurately and easily cross the boundaries of the design hierarchy so that high-level faults can be precisely related to low-level problems. Recent experiments with real ICs have shown that tests based on traditional fault models cannot provide adequate coverage of manufacturing defects. A hierarchical fault model should have the capability to model various types of faulty behaviors.

A new approach, the fault tuple (pronounced "too-pull"), can be used to create true, hierarchical test tools such as fault simulators and automatic test pattern generators. Fault tuples are simple 3-element conditions for a signal, its value, and clock cycle constraint. Fault tuples can be designed at any level of the design hierarchy or across multiple levels without loss of information. The objective of this research proposal is the development of true, hierarchical fault simulation and ATPG tools for SoCs based on the fault tuple modeling mechanism.

There are enormous demands for efficient and high performance video compression for use in multimedia communications such as for videoconferencing, digital TV, virtual reality and telemedicine, etc. A new standard, MPEG-4, has been developed for object- or content-based compression in multimedia applications where each frame of a video sequence is segmented into a set of arbitrarily shaped objects to form video object planes, and each object is to be encoded utilizing the information of its texture, shape and motion vectors. Motion estimation and motion compensation techniques must be modified for MPEG-4 to provide functionalities such as content-based interactivity and scalability. In this project, the team will develop a highly parallel shift invariant biorthogonal wavelet transform chip and a parallel VLSI architecture for wavelet-based motion estimation for MPEG-4 video compression.

The wireless telecommunications industry is now planning deployment of a third-generation of mobile systems driven by the need for greater bandwidth in anticipation of growing demand for voice services and of an emerging market for broadband multimedia services. Unfortunately, existing video compression standards, developed for relatively benign, nearly error-free environments, cannot be directly applied in the more hostile communications environment experienced by mobile systems. Therefore, the proposed research is focused on designs for robust and error-resilient multimedia transmission over 3G and beyond wireless fading channels. To achieve this, the team plans to deliver:

Many HTTP requests arriving at a Web server are static, of the form "Get me a file." For such requests, the size of the request (i.e. the time required to service the request) is well-approximated by the size of the file, which is well-known to the server. The project team will use the knowledge of the size of the request to affect the order in which requests are serviced by a busy Web server servicing thousands of requests concurrently. Traditionally, these requests are time-shared, with each request receiving a fair share of the Web server resources. The unfair scheduling, in which priority is given to short requests, or those requests which have short remaining time, is in accordance with SRPT (shortest-remaining-processing-time-first). Initially, SRPT-like scheduling sounds like it will starve the large requests. However, new theoretical results and preliminary implementation results show that in fact, for Web-type workloads, all requests, including the largest ones, will benefit from SRPT-like scheduling. The research project proposed is a proof-of-concept implementation of SRPT-like scheduling for a standard Web server (Apache), running on a standard operating system (Linux 2.2), under a realistic Web workload.