CS631 Paper Response Form
Paper Title: Psychophysics and modern digital audio technology
Author(s): A.J.M. Houtsma
Main Point(s):
The paper explains how psychophysical knowledge about the human auditory system
is used in the MUSICAM technique to reduce the bitrate of CD sound from 1.4 megabits per second to 353,000 bits without perceivable difference in sound quality. The fact that the human auditory system cannot perceive sounds that are masked by other sounds is used to optimize quantization such that the resulting
quantization noise is just masked by the sound signal and therefore not audible. In order to adapt the level of quantization noise to the shape of the sound
signal, the signal is split into 32 filter bands that are quantized separately
and coded with a dynamically allocated number of bits. Perception experiments
reveal that for an average bit rate of about 4 bits per sample (instead of 16),
the resulting sound cannot be distinguished perceptually from the original.
Possible Use(s):
By reducing the bitrate, storage space as well as transmission
bandwidth can be reduced. That's useful for transmitting radio in
CD quality, shipping music over the net, and maybe also for
using the net as telephone (however that's called). All these
video conferencing applications probably also benefit from
reduced bitrate.
Extensions:
Human voice (for telephone calls, video conferencing ect.) has
a much more restricted bandwidth than arbitrary pieces of music
and these applications also require much lower quality than hi-fi systems.
It would be interesting to see how much the bitrate can be reduced for
these applications with the same basic idea.
CS631 home page