Self-Organizing Maps and Learning Vector Quantization forFeature Sequences

  • Authors:
  • Panu Somervuo;Teuvo Kohonen

  • Affiliations:
  • Helsinki University of Technology, Neural Networks Research Centre, P.O. Box 2200, FIN-02015-HUT, Finland;Helsinki University of Technology, Neural Networks Research Centre, P.O. Box 2200, FIN-02015-HUT, Finland

  • Venue:
  • Neural Processing Letters
  • Year:
  • 1999

Quantified Score

Hi-index 0.00

Visualization

Abstract

The Self-Organizing Map (SOM) and LearningVector Quantization (LVQ) algorithms are constructed in this work forvariable-length and warped feature sequences. The novelty is to associate an entire feature vector sequence,instead of a single feature vector, as a model with each SOM node.Dynamic time warping is used to obtain time-normalizeddistances between sequences with different lengths. Starting withrandom initialization, ordered feature sequence maps then ensue, andLearning Vector Quantization can be used to fine tune the prototypesequences for optimal class separation. The resulting SOM models, theprototype sequences, can then be used for the recognition as well assynthesis of patterns. Good results have been obtained inspeaker-independent speech recognition.