MULTI-MICROPHONE TALKER SORTING DEVICE, METHOD, AND PROGRAM

摘要

A first estimation means estimates the number and directions of sound sources, on the basis of vectors having input signals as components, said input signals being audio from each sound source that is gathered by a plurality of microphones and converted to a time frequency region. A signal calibration means specifies the sound source nearest to each microphone on the basis of the number and directions of the sound sources as estimated by the first estimation means, computes the volume that is inputted into each microphone based on the sound signal emitted from each sound source, and calibrates the input signals such that the volumes are uniform. A second estimation means estimates the number and directions of the sound sources on the basis of the calibrated input signals, and sorts talkers on the basis of the estimated sound source directions.