Parameter domain loudness estimation in parametric audio object coding

Parametric audio object coding employs principles of informed source separation for obtaining object reconstructions from the mixture signal used in the transport enabling flexible output signal rendering into output scenes unknown at the encoder. Information of the object level in the rendered output is important for loudness and dynamic range control applications, e.g., in broadcast. This paper proposes a method for estimating the object level in an arbitrary output scene based on the downmix signal level that is then projected through the combined un-mixing and rendering matrix. This avoids explicit reconstruction of the objects only for the level estimation offering computational complexity savings. In the evaluations, the proposed method shows a high estimation accuracy with a root-mean squared error of 0.26 LUFS (loudness units relative to full scale) compared to 3.7 L UFS of the baseline with object reconstructions.