New clustering algorithms applied to speaker independent isolated word recognition

This paper adresses speaker independent isolated word recognition (SIIWR) performed by an approach which uses multiple template references. In order to find the references we propose to use clustering algorithms based on criterion functions. After the theoretical description of these criterion based clustering algorithms applied to SIIWR, we give the results of tests performed with an isolated word recognizer showing that they perform better than the algorithms previously used in SIIWR.