Phase-Optimized K-SVD for Signal Extraction from Underdetermined Multichannel Sparse Mixtures
This addresses the challenging problem of ego-noise reduction for robot audition, which is incremental as it builds on existing dictionary-based methods.
The paper tackles the problem of extracting signals from underdetermined multichannel sparse mixtures, specifically for ego-noise reduction in robot audition, by proposing a phase-optimized K-SVD method that jointly estimates a dictionary, sparse activations, and source phases, demonstrating superiority over conventional techniques in real-room recordings.
We propose a novel sparse representation for heavily underdetermined multichannel sound mixtures, i.e., with much more sources than microphones. The proposed approach operates in the complex Fourier domain, thus preserving spatial characteristics carried by phase differences. We derive a generalization of K-SVD which jointly estimates a dictionary capturing both spectral and spatial features, a sparse activation matrix, and all instantaneous source phases from a set of signal examples. The dictionary can then be used to extract the learned signal from a new input mixture. The method is applied to the challenging problem of ego-noise reduction for robot audition. We demonstrate its superiority relative to conventional dictionary-based techniques using recordings made in a real room.