Second European Conference on Speech Communication and Technology

Genova, Italy
September 24-26, 1991


A Neural Net Model for Vector Quantization

Fisseha Mekuria, Tore FjÖllbrant

Department of Electrical Engineering, University of Linköping, Linköping, Sweden

In this paper a neural net model for the tree searched vector quantizer is described. The design of the codebook containing the optimal number of representative vectors, incorporates a tree structure with a modified perceptron neural at each node. The proposed algorithm gives reduction in computation requirement compared to the usual full search and tree-search algorithms and requires less memory than the tree search algorithm. Keywords: Vector Quantization, Percetron neural-net, tree search, codebook

Full Paper

Bibliographic reference.  Mekuria, Fisseha / FjÖllbrant, Tore (1991): "A neural net model for vector quantization", In EUROSPEECH-1991, 1019-1022.