Figure 2
From: Very Low Rate Scalable Speech Coding through Classified Embedded Matrix Quantization
The graph for codewords at depth 
. Arc value 
is determined based on the encountered distortion resulting from merging 
th and 
th codewords at depth 
.