Spatial Audio Coherence Signalling for Low-Bit-Rate Codecs

The patent extracts only the essential spatial parameters from the full covariance matrix. Specifically, it identifies and transmits only the dominant eigenvalues and eigenvectors that capture the most significant spatial information, discarding redundant data. This extraction approach maintains spatial reproduction accuracy while dramatically reducing metadata size for low-bit-rate coding.

The patent applies local quality by differentiating the treatment of different spatial parameters. Rather than uniformly encoding all covariance matrix elements, it applies higher precision to dominant spatial components (those with significant eigenvalues) and lower or no precision to minor components. This selective precision approach optimizes the balance between spatial accuracy and data reduction.

The patent segments the spatial parameter encoding into distinct components: direction parameters, energy ratio parameters, and coherence parameters. Each segment is processed and encoded independently using optimized methods suitable for its characteristics. This segmentation reduces overall system complexity by breaking down the complex task of full covariance matrix encoding into manageable, specialized sub-tasks.

The patent transforms the original covariance matrix representation into a different parameter space consisting of direction vectors, energy ratios, and coherence values. This parameter transformation simplifies the encoding process by converting complex matrix operations into more manageable vector and scalar operations, reducing computational complexity while preserving spatial information.

The patent introduces coherence parameters as intermediary representations that bridge the full covariance matrix and its reduced form. These coherence parameters specifically capture the spatial relationship and correlation information between different audio channels, serving as a compact mediator that preserves essential spatial coherence properties while enabling low-bit-rate transmission.

The patent performs preliminary analysis of the covariance matrix to identify and extract coherence information before the actual encoding process. By pre-processing the spatial data to isolate and emphasize coherence-related parameters, the system ensures that this critical information is preserved in the reduced representation, preventing information loss during compression.