Thai Speech Keyword Spotting using Heterogeneous Acoustic Modeling

This paper illustrates the use of acoustic modeling of spoken dialog systems. One major problem with using speech three different structures, including syllables, fillers and recognizer for delivering sequence of words to other modules keywords, for keyword spotting. Filler models and syllable in spoken dialog systems is its robustness, especially, when the models are applied to capture out-of-vocabulary words, while input speech is spontaneous and degraded with noise in the keyword models extract significant words from speech operating environment. In such a case, it is very difficult to utterances. Grammatical details are utilized with syllable models correctly recognize all the words that a user says with typical to add extra domain constraints. This improves the system's speech recognizers, requiring speech to be produced in fix ability to detect non-keyword vocabularies. Filler models syntaxes, available nowadays. Another problem is the presence associating with syllable mode...

• Type: Conference
• Year: 2007
• Where: RIVF
• Authors: S. Tangruamsub, Proadpran Punyabukkana, Atiwong Suchato
• Keywords: Filler Models, Internet Technology, Keyword Spotting, RIVF 2007, Spoken Dialog Systems