A Source Separation Approach for the Combined SBA Signals in the Joint Representation of OBA and SBA

There are three methods to implement spatial audio, which include channel-based, object-based and scene-based formats. In fact, utilizing a combination of multiple audio formats specific to certain scenes can heighten the level of personalization and immersion for the audience. Specifically, the joint representation of object-based audio and scene-based audio plays a crucial role in spatial audio. However, a significant challenge lies in the fact that the scene-based audio stems from simultaneous scene recording and thus includes all object-based audio components. To tackle this issue, this paper proposes a source separation approach to distinguish the scene-based audio component from the combined signal. This approach operates on the premise that each audio source is independent from the others and can deal with both stationary and moving audio objects. The experimental results show that the approach can effectively separate the scene-based audio component from the combined signal in first-order Ambisonics signals (FOA) or higher-order Ambisonics signals (HOA).