If we knew what it was we were doing, it would not be called research, would it? -- A. Einstein

Research interests

Self-Organizing Map (SOM) and Learning Vector Quantization (LVQ) algorithms and their applications in various practical problems.

Using SOMs and LVQ with Hidden Markov Models (HMMs) in speech recognition.


In my Master's thesis "Combinations of Learning Vector Quantization and Self-Organizing Maps with hidden Markov models in speech recognition", I present my studies in using SOMs and LVQ to train continuous observation density HMMs for the phoneme recognition system of the Laboratory of Information and Computer science.

The Licentiate's thesis "Application of Learning Vector Quantization and Self-Organizing Maps for training continuous density and semicontinuous Markov models" report the advanced studies around SOMs and LVQ with CDHMMs and SCHMMs.

My Doctor's thesis "Using Self-Organizing Maps and Learning Vector Quantization for Mixture Density Hidden Markov Models" is an introduction and collection of articles published in 1993 - 1997.


My published articles present some results from special projects in speech recognition and research to combine SOMs and LVQ with continuous density HMMs and semi-continuous HMMs. Also articles about corrective tuning by LVQ and hybrid LVQ-Viterbi training have been published. Click here for the complete list of my publications including also the new ones.

Annual Reports

1991, 1992, 1993, 1994, 1995, Triennal 1994 - 1996: part A, part B,


Our Research Staff
Mikko Kurimo <Mikko.Kurimo@hut.fi>
Last modified: Wed Feb 18 15:15:32 1998