The –Buzo–Linde–Buzo–Gray algorithm (introduced by Yoseph Linde, Andrés Buzo and Robert M. Gray in 1980) is a vector quantization algorithm to derive a good codebook.
It is similar to the k-means method in data clustering.
The algorithm
At each iteration, each vector is split into two new vectors.
A initial state: centroid of the training sequence; B initial estimation #1: code book of size 2; C final estimation after LGA: Optimal code book with 2 vectors; D initial estimation #2: code book of size 4; E final estimation after LGA: Optimal code book with 4 vectors; The final two code vectors are splitted into four and the process is repeated until the desired number of code vector is obtained.
References
The original paper describing the algorithm, as an extension to Lloyd's algorithm: