THREE DIMENSIONAL SOUND REPRODUCTION SYSTEM AND METHOD USING A BINAURAL DATA WITH MACHINE LEARNING

摘要

A stereophonic sound reproduction system using binaural data and machine learning and a method thereof are disclosed. A sound characteristic variation learning method of a stereophonic sound reproduction system using binaural data and machine learning includes the steps of: setting one sound source among sound sources in a binaural database generated by recording the sound sources with different azimuths and altitude angles as a reference sound source; extracting a sound feature of each of the reference sound source and the remaining sound sources excluding the reference sound source among the sound sources; and learning the sound characteristic variation according to the azimuth or altitude angle by using a difference between the sound characteristic of the reference sound source and the sound characteristic of each of the remaining sound sources. Accordingly, the present invention can learn the sound characteristic variation in consideration of a diffraction effect due to a right ear and a left ear of a user.