An Automatic Acquisition of Acoustical Units for Speech Recognition Based on Hidden Markov Network

We have proposed a method to acquire speech recognition units automatically. From the experimental results, vowel-consonant type acoustic segments were obtained.