Spatial enhancement for immersive stereo audio applications

In this work, we propose a stereo recording spatial enhancement technique which retains the original panning/source location, proportionally mapped into the perceived expanded sound stage. The technique uses a time-frequency domain metric for retrieving the panning coefficients applied during the initial stereo mixing. Panning information is also used for separating the original single channel audio streams and finally for synthesizing the expanded sound field using binaural processing. The technique is mainly intended for playback applications over short-distant loudspeaker setups or headphones and, in the context of this work, it is subjectively evaluated in terms of the perceived sound field expansion and the sound-source spatial distribution accuracy.