Granted Patent B2

US 12,267,534 · App. 17/888,816 · Granted Apr 1, 2025

Inventors: Ramesh Naidu Bheemarasetty (Little Elm, TX); Madhavan Sivathanu Pillai (Little Elm, TX)

Assignee: Verizon Patent and Licensing Inc.

H04N21/23424H04N21/2385H04N21/4662

Abstract

Claims (86)

1. A method, comprising:

receiving a prediction input characterizing utilization of a plurality of content stream utilization channels (CSUCs);

predicting automatically one or more schedule parameters based on the prediction input via one or more of multiple different prediction models obtained via machine learning based on training data that is generated based on historic data related to content delivery via the plurality of CSUCs, wherein the one or more schedule parameters comprise a CSUC occupancy prediction associated with a time frame in a geographical region and/or in an adapter, wherein the historic data is grouped in accordance with some criterion determined based on operational modes in which the multiple different prediction models are to operate, and wherein the multiple different prediction models are trained based on different groups of historic data, respectively;

identifying, based on the one or more schedule parameters, at least one insertion opportunity with respect to at least one CSUC;

generating, for each of the at least one insertion opportunity, a corresponding insertion schedule specifying insertion of a content stream into a selected CSUC at an insertion time; and

updating the training data based on content insertion schedules generated according to predictions made based on the one or more of multiple different prediction models, wherein the one or more of multiple different prediction models are updated based on the updated training data.

2. The method of claim 1 , wherein the one or more schedule parameters relate to candidate insertion opportunities and further include at least one of:

a start time prediction;

and

a collision prediction.

3. The method of claim 1 , wherein the multiple different prediction models include:

a start time prediction model;

a CSUC occupancy prediction model; and

a collision prediction model.

4. The method of claim 1 , wherein:

the historic data is collected by one or more sources;

the grouped historic data is processed to generate training data, which is used to learn the multiple different prediction models in an iterative process; and

the multiple different prediction models are generated based on the learning result.

5. The method of claim 4 , wherein the grouped historic data is processed via one or more of:

removing each outlier from the historic data;

normalizing the historic data;

encoding the historic data; and

scaling the historic data.

6. The method of claim 2 , wherein the step of identifying comprises:

excluding each candidate insertion opportunity that is within a collision window defined based on a collision prediction to generate a remaining set of candidate insertion opportunities;

for each start time prediction in the remaining set of candidate insertion opportunities,

determining CSUC availability with respect to the start time prediction based on CSUC occupancy predictions in the remaining set of candidate insertion opportunities, and

selecting an insertion opportunity with respect to the CSUC availability in accordance with some predetermined criteria.

7. The method of claim 1 , further comprising inserting, according to each insertion schedule, a specified content stream into a selected CSUC at a designated insertion time.

8. A machine readable and non-transitory medium having information recorded thereon, wherein, when the information is read by the machine, causes the machine to perform the following:

receiving a prediction input characterizing utilization of a plurality of content stream utilization channels (CSUCs);

predicting automatically one or more schedule parameters based on the prediction input via one or more of multiple different prediction models obtained via machine learning based on training data that is generated based on historic data related to content delivery via the plurality of CSUCs, wherein the one or more schedule parameters comprise a CSUC occupancy prediction associated with a time frame in a geographical region and/or in an adapter, wherein the historic data is grouped in accordance with some criterion determined based on operational modes in which the multiple different prediction models are to operate, and wherein the multiple different prediction models are trained based on different groups of historic data, respectively;

identifying, based on the one or more schedule parameters, at least one insertion opportunity with respect to at least one CSUC;

generating, for each of the at least one insertion opportunity, a corresponding insertion schedule specifying insertion of a content stream into a selected CSUC at an insertion time; and

updating the training data based on content insertion schedules generated according to predictions made based on the one or more of multiple different prediction models, wherein the one or more of multiple different prediction models are updated based on the updated training data.

9. The medium of claim 8 , wherein the one or more schedule parameters relate to candidate insertion opportunities and further include at least one of:

a start time prediction;

and

a collision prediction.

10. The medium of claim 8 , wherein the multiple different prediction models include:

a start time prediction model;

a CSUC occupancy prediction model; and

a collision prediction model.

11. The medium of claim 8 , wherein:

the historic data is collected by one or more sources;

the grouped historic data is processed to generate training data which is used to learn the multiple different prediction models in an iterative process; and

the multiple different prediction models are generated based on the learning result.

12. The medium of claim 11 , wherein the grouped data is processed via one or more of:

removing each outlier from the historic data;

normalizing the historic data;

encoding the historic data; and

scaling the historic data.

13. The medium of claim 9 , wherein the step of identifying comprises:

excluding each candidate insertion opportunity that is within a collision window defined based on a collision prediction to generate a remaining set of candidate insertion opportunities;

for each start time prediction in the remaining set of candidate insertion opportunities,

determining CSUC availability with respect to the start time prediction based on CSUC occupancy predictions in the remaining set of candidate insertion opportunities, and

selecting an insertion opportunity with respect to the CSUC availability in accordance with some predetermined criteria.

14. The medium of claim 8 , wherein the information, when read by the machine, further causes the machine to perform inserting, according to each insertion schedule, a specified content stream into a selected CSUC at a designated insertion time.

15. A system, comprising:

a model-based schedule parameter predictor implemented on a processor and configured for

receiving a prediction input characterizing utilization of a plurality of content stream utilization channels (CSUCs), and

predicting automatically one or more schedule parameters based on the prediction input via one or more of multiple different prediction models obtained via machine learning based on training data that is generated based on historic data related to content delivery via the plurality of CSUCs, wherein the one or more schedule parameters comprise a CSUC occupancy prediction associated with a time frame in a geographical region and/or in an adapter, wherein the historic data is grouped in accordance with some criterion determined based on operational modes in which the multiple different prediction models are to operate, and wherein the multiple different prediction models are trained based on different groups of historic data, respectively;

a content insertion schedule generator implemented on a processor and configured for

identifying, based on the one or more schedule parameters, at least one insertion opportunity with respect to at least one CSUC, and

generating, for each of the at least one insertion opportunity, a corresponding insertion schedule specifying insertion of a content stream into a selected CSUC at an insertion time; and

a dynamic data collector implemented on a processor and configured for

updating the training data based on content insertion schedules generated according to predictions made based on the one or more of multiple different prediction models, wherein the one or more of multiple different prediction models are updated based on the updated training data.

16. The system of claim 15 , wherein

the one or more schedule parameters relate to candidate insertion opportunities and further include at least one of a start time prediction and a collision prediction; and

the multiple different prediction models include a start time prediction model, a CSUC occupancy prediction model, and a collision prediction model.

17. The system of claim 15 , further comprising a machine learning engine implemented on a processor and configured for obtaining, via machine learning, the multiple different prediction models based on training data derived by:

receiving grouping the historic data collected by one or more sources;

processing the grouped historic data to generate the training data;

learning the multiple different prediction models based on the training data in an iterative process; and

generating the multiple different prediction models based on the learning result.

18. The system of claim 17 , wherein the machine learning engine comprises one or more of:

a first unit implemented on a processor and configured for removing each outlier from the historic data;

a second unit implemented on a processor and configured for normalizing the historic data;

a third unit implemented on a processor and configured for encoding the historic data; and

a fourth unit implemented on a processor and configured for scaling the historic data.

19. The system of claim 15 , wherein the content insertion schedule generator is further configured for identifying the at least one insertion opportunity by:

excluding each candidate insertion opportunity that is within a collision window defined based on a collision prediction to generate a remaining set of candidate insertion opportunities;

for each start time prediction in the remaining set of candidate insertion opportunities,

determining CSUC availability with respect to the start time prediction based on CSUC occupancy predictions in the remaining set of candidate insertion opportunities, and

selecting an insertion opportunity with respect to the CSUC availability in accordance with some predetermined criteria.

20. The system of claim 15 , further comprising a schedule-based content insertion unit implemented on a processor and configured for inserting, according to each insertion schedule, a specified content stream into a selected CSUC at a designated insertion time.

Assignments (1)

Continuity (1)

References Cited (16)

Cited By (1)

US 12,657,144