US 11,705,148 B2
Adaptive coefficients and samples elimination for circular convolution
Ziyad Ibrahim, Redmond, WA (US); Laxmi Narsimha Rao Kakulamarri, Redmond, WA (US); and Andrew Yi Jiang, Seattle, WA (US)
Assigned to MICROSOFT TECHNOLOGY LICENSING, LLC, Redmond, WA (US)
Filed by MICROSOFT TECHNOLOGY LICENSING, LLC, Redmond, WA (US)
Filed on Oct. 13, 2021, as Appl. No. 17/500,893.
Claims priority of provisional application 63/209,942, filed on Jun. 11, 2021.
Prior Publication US 2022/0399029 A1, Dec. 15, 2022
Int. Cl. G10L 21/0332 (2013.01); G10L 19/022 (2013.01); G10L 21/007 (2013.01); H04S 7/00 (2006.01)
CPC G10L 21/0332 (2013.01) [G10L 19/022 (2013.01); G10L 21/007 (2013.01); H04S 7/301 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method for efficient real-time audio processing, the method comprising:
receiving an audio signal block comprising a first plurality of frequency samples ordered from a lowest frequency to a highest frequency;
identifying an impulse response block to convolve with the audio signal block to generate an output audio signal, wherein the impulse response block comprises a second plurality of frequency samples ordered from the lowest frequency to the highest frequency;
identifying a first threshold frequency above which frequency samples of the first plurality of frequency samples would contribute less than a threshold amount of volume to the output audio signal;
identifying a second threshold frequency above which frequency samples of the second plurality of frequency samples would contribute less than the threshold amount of volume to the output audio signal;
copying, to an audio processing device, frequency samples of the second plurality of frequency samples that are less than the second threshold frequency;
identifying as a maximum frequency a minimum of the first threshold frequency and the second threshold frequency;
performing a convolution, from the lowest frequency to the maximum threshold frequency, on frequency samples from the first plurality of frequency samples with frequency samples from the second plurality of frequency samples; and
generating the output audio signal based on a result of the convolution.