SPEECH PROCESSING DEVICE, METHOD, AND PROGRAM

摘要

PROBLEM TO BE SOLVED: To allow an arbitrary content to be identified with higher accuracy.SOLUTION: An input signal processing unit extracts, from an input signal of a content as an identification object, a sound feature quantity IA1 indicative of a likelihood that the signal in a time-frequency domain is a sine wave, and a sound feature quantity IA2 indicative of individuality of the signal. A reference signal processing unit extracts a sound feature quantity RA1 and a sound feature quantity RA2 corresponding to the sound feature quantity IA1 and the sound feature quantity IA2 from a reference signal of an original content. A matching processing unit calculates a mask pattern from the sound feature quantity IA1 and the sound feature quantity RA1 and calculates similarity between these sound feature quantities. The matching processing unit calculates similarity between the input signal and the reference signal on the basis of the sound feature quantity IA2 and the sound feature quantity RA2, thereby identifying the content. This technique is applicable to a speech processing device.