Method and apparatus for recognizing the uttered words in a speech signal

摘要: In the recognition of coherent speech, language models are favourably used to increase reliability recognition, which models, for example, take into account probabilities word combinations, especially pairs. For this purpose, a model value corresponding probability is added at boundaries between words. several methods, when vocabulary built up from phonemes in shape tree, it not known start continuation hypothesis after end will actually follow, so that cannot be taken until next word. Measures given achieving such manner as far possible optimal preceding or sequence without necessity constructing copy searching tree each and every simultaneously ending sequence.