14.arch/JW.lazzaro
.ls 2
.na
.LP
Analog VLSI Sensory Processing

John Lazzaro

(Professor J. Wawrzynek)

AT&T Bell Laboratories, 
(NSF) MIP-89-58568, and 
(ONR) URI-N00014-92-J-1672

Our work centers on building practical computing systems that
use analog integrated circuit models of biological processing as
components. Our current focus is on auditory processing, with the
goal of producing useful devices for speech recognition, music
processing, and other auditory applications.

Complementary research on the use of these analog models in
digital computing systems is also needed. Simplistic approaches to
these problems can eliminate the benefits of analog processing. The
low power consumption, compact circuits, and clean representations of
analog auditory models are only relevant if the system integration
approach complements and preserves these characteristics. Our research
centers on developing appropriate circuit, architecture, and software
integration technologies for these models. 

Integrating these analog elements into a digital computer
system requires a careful approach. The interface methods must
complement and preserve the low power consumption, compact
circuits, and clean representations of analog auditory models. Our
research centers on developing appropriate circuit, architecture, and
software integration technologies for these models.

Another major practical issue is the generation of
digitally programmable analog parameters, to control the behavior of
the analog processing. In collaboration with Alan Kramer, we have
developed a novel method for using floating-gate tunneling techniques
for non-volatile analog storage of parameters. This method does not
need special VLSI processing steps; it allows for digital control of
the analog parameters. 

We have recently fabricated an analog auditory model that uses
these two technologies. The chip is a 122-channel auditory model of
spectral shape. Sound input to the chip is an analog signal, while the
output from the chip is digital; in this sense the chip is a
special-purpose, low-power analog-to-digital converter for auditory
applications. We are currently evaluating the suitability of the chip
as a processor for music and speech applications.

[1]	J. P. Lazzaro, J. Wawrzynek, M. Mahowald, M. Sivilotti, 
and D. Gillespie, "Silicon Auditory Processors as Computer
Peripherals," IEEE J. Neural Networks, Vol. 4, No. 3, 1993,
pp. 523-528.

[2]	J. P. Lazzaro, "A Silicon Model of an Auditory Neural
Representation of Spectral Shape," IEEE J. Solid State Circuits,
Vol. 26, 1991, pp. 772-777.