Next  |  Prev  |  Up  |  Top  |  JOS Index  |  JOS Pubs  |  JOS Home  |  Search

Coefficient Clustering - Frequency Domain

The first thing to be done in a more advanced approach is to cluster the masking thresholds in the frequency, so that quantization step sizes do not take too many bits. This is often done in critical bands. This coder uses 24 slightly adjusted critical bands, in each of which the amplitude mean of the masking threshold is used as quantization step size. The bands were adjusted to better use the total clipping which often occurs at high frequencies, due to the ATH.



Band First bin Last bin Coefficient ratio
1 0 1 0.874486
2 2 3 0.810296
3 4 5 0.831361
4 6 7 0.835856
5 8 9 0.894511
6 10 11 0.938870
7 12 13 1.000000
8 14 15 1.000000
9 16 18 1.000000
10 19 21 1.000000
11 22 24 1.000000
12 25 28 1.000000
13 29 33 0.932482
14 34 38 0.962853
15 39 45 0.839628
16 46 53 0.789081
17 54 64 0.685002
18 65 77 0.684814
19 78 91 0.712828
20 92 108 0.683237
21 109 128 0.672980
22 129 149 0.681761
23 150 170 0.818602
24 171 255 0.894698


The table shows the clustering of the frequency bins into bands. The ``Coefficient ratio'' column is described below for stationary coding mode.


Next  |  Prev  |  Up  |  Top  |  JOS Index  |  JOS Pubs  |  JOS Home  |  Search

Download bosse.pdf

``An Experimental High Fidelity Perceptual Audio Coder'', by Bosse Lincoln<bosse@ccrma.stanford.edu>, (Final Project, Music 420, Winter '97-'98).
Copyright © 2006-01-03 by Bosse Lincoln<bosse@ccrma.stanford.edu>
Center for Computer Research in Music and Acoustics (CCRMA),   Stanford University
CCRMA  [Automatic-links disclaimer]