Subband Audio Filter Parameterization for Low-Complexity Binaural Rendering

The patent replaces time-domain convolution operations with frequency-domain filtering operations. Specifically, it substitutes the mechanical computation of long-time convolution (which requires enormous computational resources) with frequency-domain filter applications that achieve the same acoustic effect with much lower computational complexity. The BRIR filter is transformed into frequency domain through FFT, and the filtering operation is performed by simple multiplication in frequency domain rather than time-domain convolution.

The patent changes the domain parameter from time domain to frequency domain. By applying FFT to transform the BRIR impulse response from time domain to frequency domain, the system changes the operational parameter space. This transformation allows the same filtering function to be achieved with different computational characteristics - specifically, replacing O(N*M) time-domain convolution complexity with O(N log N) frequency-domain processing complexity.

The patent merges the binaural rendering filtering operation with the existing QMF synthesis operation that is already required by the audio codec. Instead of performing separate time-domain convolution and then separate QMF synthesis, the invention integrates both operations into a single frequency-domain filtering step. The BRIR frequency response is applied directly to the QMF subband signals, combining what would otherwise be two distinct processing stages into one unified operation.

The patent makes the frequency-domain filter serve multiple functions simultaneously. The same frequency-domain filter coefficients are used both for implementing the binaural spatial rendering and for performing the QMF synthesis required by the audio decoding standard. This multi-functionality eliminates redundant operations and reduces overall system complexity while maintaining both spatial audio quality and decoding compatibility.