Multichannel audio resynthesis based on a generalized Gaussian mixture model and cepstral smoothing

Multichannel audio is an emerging technology with continuously increasing applications. Audio reproduction through multiple channels has the advantage of recreating the acoustic scene with unprecedented fidelity and of immersing the listener in an acoustic environment that is virtually indistinguishable from reality. However, one of the greatest challenges of this scheme is its high transmission requirements especially since accurate rendering through as many possible channels is the main purpose. This paper follows previous techniques on spectral conversion and a recently introduced concept called audio resynthesis. In audio resynthesis, a reference channel is transmitted and then used to recreate the remaining channels at the receiver. An alternative approach to audio resynthesis is presented based on the generalized Gaussian mixture model. This model incorporates most of the standard mixtures (Laplace, Gaussian etc) but this flexibility comes with high structural complexity due to the increased number of model parameters. A scheme is presented here that bypasses this issue and avoids the use of the expectation-maximization (EM) algorithm. A smoothing technique is also introduced which optimizes the performance during the spectral conversion stage and significantly improves the resynthesis results.