APPARATUS AND METHOD FOR EXTRACTING FEATURES FOR SPEECH RECOGNITION

摘要

According to the present invention, an apparatus for extracting features for a voice recognition comprises: a frame forming unit which divides an inputted voice signal into frame units of a predetermined size; a static features extracting unit which extracts a static feature vector to each frame of the voice signal; a dynamic feature extracting unit which extracts a dynamic feature vector indicating a change with time of the extracted static feature vector by using a basis function or a basis vector; and a feature vector combining unit which combines the extracted static feature vector and the extracted dynamic feature vector and forms a feature vector stream.