Convention Paper
Partial Unmixing for Personalized Audio
AES Convention 117 · Paper 6295 · October 2004
Abstract
Immersive audio for interactive gaming is necessarily processed and mixed in real time as it is being rendered on the game audio playback platform. It is generally assumed that music and movie soundtracks require no comparable processing during playback because listeners typically provide no real-time input that might affect the final rendering. In reality, pre-packaged audio is being delivered to music and movie playback platforms in increasingly diverse forms. The result is that mismatches between the spatial audio format, bit depth, and frequency range of the content and of the playback system pose an emerging problem for which sophisticated playback processing may be an appropriate response. This paper presents a formal statement of the mismatch problem and proposes a unified solution using frequency-domain processing to perform "partial unmixing" of the pre-packaged content. Lastly, we show how this can enable a new music/movie listening experience rooted in the concept of "personalized audio."
