VOICE QUERY EXTENSION METHOD AND SYSTEM

摘要

A method and a system for extending a voice search word are provided to raise a recognition hit ratio by recognizing unregistered words, which is not registered to a native language dictionary, with a multi-language dictionary. A signal processor detects a section of a voice user from an input signal and extracts a feature vector from the voice section. A phoneme decoder(110) converts the feature vector into at least one phoneme sequence. A word decoder(140) extracts and selects a row of words, which have linguistic meaning, as a search word by matching the phoneme sequence with the words registered in a word dictionary. A multi-language search extender(180) converts the selected search word into the search word comprising first language by using a phoneme/grapheme conversion model when the selected search word is not the first language.