US 12,462,261 B2
Optimizing carbon emissions from streaming platforms with artificial intelligence based model
Divgian Sidhu, Mohali (IN); Ayush Jain, Lucknow (IN); Smitkumar Narotambhai Marvaniya, Bangalore (IN); and Sujoy Kumar Roy Chowdhury, Kolkata (IN)
Assigned to INTERNATIONAL BUSINESS MACHINES CORPORATION, Armonk, NY (US)
Filed by INTERNATIONAL BUSINESS MACHINES CORPORATION, Armonk, NY (US)
Filed on Aug. 19, 2021, as Appl. No. 17/406,722.
Prior Publication US 2023/0059038 A1, Feb. 23, 2023
Int. Cl. H04N 21/2343 (2011.01); G06F 11/34 (2006.01); G06Q 30/018 (2023.01); G06Q 30/0282 (2023.01)
CPC G06Q 30/018 (2013.01) [G06F 11/3428 (2013.01); G06Q 30/0282 (2013.01); H04N 21/2343 (2013.01)] 18 Claims
OG exemplary drawing
 
6. A system for optimizing streaming media comprising:
a hardware processor; and
a memory that stores a computer program product, which, when executed by the hardware processor, causes the hardware processor to:
analyze a streaming history of a streaming subscription to determine a historical carbon footprint;
train a neural network to forecast an upcoming carbon footprint based on the historical carbon footprint by matching search items extracted from user requirements and source code stored in repositories to obtain a trained neural network;
present, using the trained neural network, at least one streaming plan includes a target carbon footprint relative to the upcoming carbon footprint;
receive a selected streaming plan for one of the at least one streaming plan having a target carbon emissions, wherein a streaming performance on the selected streaming plan is tracked for post streaming plan carbon emissions in carbon unit of measure;
maximize the streaming performance and minimize the carbon emissions using multi-objective optimization during streaming by dynamically modifying the streaming performance including:
divide content being broadcast into frames by performing video segmentation of the content being broadcast responsive to a temporal change meeting a threshold value;
analyze an aesthetic quality of the frames based on frame content parameters to determine a carbon output for the frames;
train a multi-modal model with historical trends of carbon emissions per calendar time periods for the streaming media being broadcast based on the carbon output for the frames to obtain a trained multi-modal model;
predict, with the trained multi-modal model, an upcoming carbon footprint as a streaming performance of content per frame for carbon units being produced from the streaming media; and
adjust, with the trained multi-modal model, the frames into carbon-aware frames by optimizing the carbon units produced by the frames based on user streaming characteristics and the aesthetic quality of the frames to match the target carbon footprint of the selected streaming plan and the upcoming carbon footprint in real time including reducing resolution for images including only black and white text, and increasing resolution for images of motion and color images.