摘要 |
<p>A car navigation system having a speech recognizing function capable of retrieving position information through speech recognition without requiring any registration of speech recognition reference pattern of a word to be recognized. The car navigation system comprises an acoustic unit recognizing means (115) which collates an input acoustic feature vector produced by acoustic analysis of the speech uttered by a user with an acoustic model of each predetermined acoustic unit of a syllable or a phoneme and outputs a related acoustic unit recognition candidate series, a word and position information registering means (118) which relates the acoustic unit recognition candidate series to position information acquired from a car navigation system body to each other and registers them in a word and position information relation dictionary, and a position information retrieving and outputting means (119) which collates the input acoustic feature vector with the acoustic characteristic vector of each acoustic unit recognition candidate series in the word and position information relation dictionary, calculates the likelihood, and outputs to the car navigation system body the position information related to the acoustic unit recognition candidate series with the likelihood not less than a predetermined threshold.</p> |