Abstract
Acoustic transients - short, impulsive bursts of acoustic energy - are a rich source of information in the natural world. Biological systems process them quickly and economically. In this article, we describe a biologically inspired analog very-large-scale integration (VLSI) architecture for real-time classification of acoustic transients. Judicious normalization of time-frequency signals allows an elegant and robust implementation of a correlation algorithm. The algorithm replaces analog-analog multiplication with binary multiplexing of analog signals. This removes the need for analog storage and analog multiplication. Simulations show that the resulting algorithm has the same out-of-sample classification performance (about 93% correct) as a template-matching algorithm based on conventional analog correlation. This development paves the way for intelligent acoustic processing in low-power applications such as cellular telephones and debit cards.
Original language | English (US) |
---|---|
Pages (from-to) | 244-252 |
Number of pages | 9 |
Journal | Johns Hopkins APL Technical Digest (Applied Physics Laboratory) |
Volume | 18 |
Issue number | 2 |
State | Published - Apr 1 1997 |
Externally published | Yes |
Keywords
- Acoustic transients
- Analog VLSI
- Matched filters
- Neural computation
- Speech recognition
ASJC Scopus subject areas
- General Engineering
- General Physics and Astronomy