Estimation of a Sound Image Flow using a Moving Image Processing Technique

Kenji Ozawa, Kodai Saito, Yuichiro Kinoshita, Satoshi Oode, and Akio Ando


Sound image, Movement, Block-matching technique, Interaural cross-correlation function, Presence


Movement of a sound image is an essential factor for perceiving the sense of presence from Audio--Visual (AV) content. This study proposes an estimation method for sound image flow, which represents the movement of a sound image as a vector. A key point of the method is that sound image flow is estimated based on binaural signals by using a moving image processing technique. First, interaural cross-correlation functions are calculated for short-duration binaural signals in multiple frequency bands. These functions are then converted into a still picture by substituting the value of the cross-correlation with the lightness of a pixel. These pictures are repeatedly produced for subsequent signal periods so that a moving picture is obtained. The block-matching technique is applied to extract moving objects in the moving picture. Finally, sound image flows are calculated. The proposed method was applied to 10 binaurally recorded sound samples, and these samples were evaluated by 10 listeners in a psychoacoustical experiment. The experimental results showed that the estimated directions of sound image flows were similar to those perceived by the listeners. Moreover, the magnitude of sound image flows corresponded well with the sense of movement.

Important Links:

Go Back