MIT Department of Electrical Engineering & Computer Science

E E C S

Adding Noise to Improve High Fi Audio Compression

Meir Feder
Tel Aviv University and MIT

Monday, March 11, 1996
4:00 PM (3:45 refreshments)
Edgerton Hall, Room 34-101

Abstract

Dithered quantization, where a small noise (dither) is added to the signal prior to quantization, was proposed by Larry Roberts more than 30 years ago. Originally it was motivated mostly by image coding problems, and justified as a tool to replace quantization bias error with variance error.

In this talk we first present several quantization and compression schemes based on Entropy Constrained Dithered Quantization (ECDQ) i.e., on dithered quantization followed by lossless (entropy) coding. We discuss the universal properties of these schemes, and present basic results regarding their attainable rate-distortion trade-off. Specifically we show the relatively small loss in performance as compared with the optimal performance given by the rate-distortion function of the source.

In the second part we discuss the principles of Hi-Fi audio compression. The most efficient usage of ECDQ includes pre- and post-filters. This structure fits audio coding principles, so ECDQ can be naturally applied to Hi-Fi audio compression. We will demonstrate results of Hi-Fi stereo compressed to 64 Kbit/sec and 32 Kbit/sec per channel, by our proposed dithered quantization system. This demonstration shows that adding noise can indeed improve the quality of compressed Hi-Fi audio.

The talk presents work done jointly with Dr. Ram Zamir (currently at U. of Calfornia Santa Barbara) and Eli Shabiro of Tel-Aviv University.

URL of this page: http://www-eecs.mit.edu/AY95-96/events/27.html
Created: Mar 7, 1996 | Modified: Jun 25, 1997
This announcement is from the MIT EECS 1995-96 archive. | Current events
To MIT EECS home page | Your comments and inquiries are welcome.