From: Very Low Rate Scalable Speech Coding through Classified Embedded Matrix Quantization
Codebook type
1st split
2nd split
3rd split
4th split
5th split
Total
Mixed voicing
10
9
8
47
Unvoiced
7
6
37