Efficiency analysis of speech recognition with hidden Markov-models

Abstract

Speech recognition is the process with which a speech recognition machine identifies the pronounced speech signals and converts them to text or other computer-processable data. By speech signals, of course, we can also mean acoustic or even visual signals (gestures, facial expressions, mouth movements). The speech recognizer I have taught, on the other hand, will take into account the acoustic signals, i.e. the speech itself, which it will convert to text. My research focuses on choosing the optimal number of states for the hidden Markov - models that form the basis for highlighting the essentials.