Analysis of Spectral Characteristics of Audio Object for Enhancement of Spatial Audio Object Coding

Abstract

This paper describes a method that can enhance the performance of spatial audio object coding (SAOC) with two-step coding structure to make a multi-object based audio service with users' interaction. In the proposed method, frequency bands and bit-rate of a residual signal can be variably decided by analysing the spectral characteristics of a specific audio object such as vocal object to efficiently handle the residual signal which is used to precisely control the specific audio object. The proposed method divided the frequency band of the specific audio object into five frequency regions including silence according to the spectral characteristics of the specific audio object. From the analysis result, most audio objects including vocal object have about more than 20 percents low frequency region compared with the original fixed frequency band to be 0 to 5.5 kHz. Consequently, the proposed method showed the high bit-rate reduction of about 28.6 percents for test items and about more than 10 percents for audio objects. Especially, since the bit-rate of the vocal object which is mainly used in KARAOKE application scenario can be greatly reduced as about 15 percents, it can be confirmed that the proposed method can enhance the original SAOC with two-step coding structure.