The present invention relates to a keyword detection system and a method of implementing a filler model for each keyword having a filler model for each keyword for improving a keyword detection capability by separately implementing a filler model for each keyword. To this end, the present invention compares the feature vector extracted from the speech signal with the stored keywords to calculate and output a similarity (Likelihood) that is close to the keyword, and the sound of each keyword for each keyword model The keyword is detected by comparing the filler model which calculates and outputs the similarity with respect to the feature vector and the similarity received from each keyword model and the similarity received from the perl model corresponding to each keyword model. It is preferable to include a similarity comparison section. Accordingly, the present invention can improve the keyword detection performance by implementing the filler model for each keyword according to the acoustic characteristics of each keyword.
展开▼