Partial Loudness in Multitrack Mixing

Partial loudness can be used as high-level, perceptually relevant metadata in the context of semantic audio, especially in multitrack mixtures or wherever masking scenarios are desired. Subjective evaluation of the partial loudness model of Glasberg and Moore on multitrack signals in the form of equal loudness matching experiment is presented. The observed results imply that the current model underrates the partial loudness perceived by the subjects. We analyze the underlying features and propose a parameter modification in the implementation of the partial loudness model that yields better compliance for musical signals.