Market customers classification using Hidden Markov Models toolkit

This paper presents a new system for pattern recognition. The system is built on Hidden Markov Models (HMMs). In this work we adapt the Hidden Markov Models Toolkit (HTK) to deal with the pattern recognition issue. HTK was originally designed for speech recognition research. Patterns are initially described by the mean of feature vectors. Those feature vectors are then converted to HTK format by adding headers and representing them in successive frames. Each one is multiplied by a Windowing function. Feature vectors are then used by HTK for training and recognition test. For experiments, we use 1600 randomly generated pattern belonging to sixteen classes of customers. Obtained results show the efficiency of the proposed approach.