Improving time-frequency sparsity for audio spatialization by time-adaptive windowing

We propose a new time-adaptive windowing technique to obtain a sparse time-frequency representation for audio signals. This transformation helps in providing better source separation from stereo mixtures for improved subsequent spatial rendering over headphones. We start with standard stereo audio recordings, transform them to a sparse representation and then estimate the mixing parameters to be used for source separation. The performance of the new representation is compared with existing methods via the accuracy of mixing parameters estimation for a test dataset of multi-speaker stereo mixtures.