US 12,406,658 B2
Method and apparatus combining separation and classification of audio signals
Aaron Steven Master, San Francisco, CA (US); Lie Lu, Dublin, CA (US); and Heidi-Maria Lehtonen, Upplands Väsby (SE)
Assigned to Dolby Laboratories Licensing Corporation, San Francisco, CA (US); and DOLBY INTERNATIONAL AB, Dublin (IE)
Appl. No. 17/921,564
Filed by Dolby Laboratories Licensing Corporation, San Francisco, CA (US); and DOLBY INTERNATIONAL AB, Dublin (IE)
PCT Filed May 3, 2021, PCT No. PCT/US2021/030506
§ 371(c)(1), (2) Date Oct. 26, 2022,
PCT Pub. No. WO2021/225978, PCT Pub. Date Nov. 11, 2021.
Claims priority of provisional application 63/019,597, filed on May 4, 2020.
Claims priority of application No. 20172644 (EP), filed on May 4, 2020.
Prior Publication US 2023/0215423 A1, Jul. 6, 2023
Int. Cl. G10L 21/0272 (2013.01); G10L 15/08 (2006.01); G06F 3/16 (2006.01); G10L 25/24 (2013.01)
CPC G10L 15/08 (2013.01) [G10L 21/0272 (2013.01); G06F 3/165 (2013.01); G10L 25/24 (2013.01)] 15 Claims
OG exemplary drawing
 
1. A computer-implemented audio separation and classification method comprising:
performing source separation on a mixture of audio signals to obtain at least one separated audio signal containing a target audio source or sources;
performing audio classification on a linear combination of the mixture of audio signals and the separated audio signal to obtain a determination whether the mixture of audio signals contains the target audio source or sources; and
time gating the separated audio signal based on at least the determination.