Intelligibility of HE-AAC Coded Japanese Words with Various Stereo Coding Modes in Virtual 3D Audio Space

摘要

In this paper, we investigated the influence of stereo coding on Japanese speech localized in virtual 3-D space. We encoded localized speech using joint stereo and parametric stereo modes within the HE-AAC encoder. First, we tested subjective quality of localized speech at various azimuths on the horizontal plane relative to the listener using the standard MUSHRA tests. We compared the encoded localized speech quality with various stereo encoding modes. The joint stereo mode showed significantly higher MUSHRA scores than the parametric stereo mode at azimuths of ±45 degrees. Next, the Japanese word intelligibility tests were conducted using the Japanese Diagnostic Rhyme Tests. Test speech was first localized at 0 and ±45 degrees and compared with localized speech with no coding. Parametric stereo-coded speech showed lower scores when localized at -45 degrees, but all other speech showed no difference between speech samples with no coding. Next, test speech was localized in front, while competing noise was localized at various angles. The two stereo coding modes with bit rates of 56, 32, and 24 kbps were tested. In most cases, these conditions show just as good intelligibility as speech with no encoding at all noise azimuths. This shows that stereo coding has almost no effect on the intelligibility in the bit rate range tested.