Phase recovery of modified spectrograms is a major issue in audio signalprocessing applications, such as source separation. This paper introduces anovel technique for estimating the phases of components in complex mixtureswithin onset frames in the Time-Frequency (TF) domain. We propose to exploitthe phase repetitions from one onset frame to another. We introduce a referencephase which characterizes a component independently of its activation times.The onset phases of a component are then modeled as the sum of this referenceand an offset which is linearly dependent on the frequency. We derive a complexmixture model within onset frames and we provide two algorithms for theestimation of the model phase parameters. The model is estimated onexperimental data and this technique is integrated into an audio sourceseparation framework. The results demonstrate that this model is a promisingtool for exploiting phase repetitions, and point out its potential forseparating overlapping components in complex mixtures.
展开▼