This research deals with the implementation of continuous-time analog VLSI chaps for processing of sensory signals. Two specific analog sensors are being developed: 1) a single-chip visual time- to-contact sensor that reports the amount of time until the chip collides with an oncoming obstacle (assuming constant relative motion), and, 2) a sound localization chip that computes the direction (in two-dimensions) of a sound source using signals from two microphones. Though these projects may appear to be unrelated, the proposed solutions using adaptive analog computation in CMOS hardware lead to many common problems that must be solved. By providing a cross-fertilization between auditory and visual processing, it is hoped that the underlying biological computational primitives and neuronal representations will be discovered. The resulting low-power analog devices have obvious applications in such areas as collision warning sensors for intelligent automobiles and improved, low-power hearing aids. The educational plan at both the undergraduate and graduate level has a strong interdisciplinary emphasis. Too often students are taught to specialize in a narrow area and are consequently unable to relate their work to an overall system objective. Teaching objectives are: 1) to stress the interdisciplinary nature of real- world systems, and 2) to introduce students to problem solving techniques for real-world signal-processing applications. In existing classes, the use of laboratories and computer experiments where students can do hands-on experimentation with concepts is emphasized. New courses at the undergraduate and graduate levels that give students more exposure to real-world experimentation in laboratories will be designed.
