SYSTEMS AND METHOD OF DETECTING TELEVISION STATE BASED ON MACHINE LEARNING

20250071376

18/452833

August 21, 2023

Server to server direct integration is provided. A system constructs an array of power values from time series data collected by a power sensor coupled with a television. The system inputs the array of power values into a model trained with machine learning based on log files of data collected from a plurality of power sensors coupled with a plurality of televisions. The system determines, based on output from the model generated with the array of power values, the television is active. The system executes, responsive to the determination that the television is active, an action to evaluate a performance of content rendered by the television.

Kantar Group Limited (New York, NY, US)

1. A system of state-based performance evaluation of televisions via machine learning, comprising: a computing system comprising one or more processors, coupled with memory, to: construct an array of power values from time series data collected by a power sensor coupled with a television; input the array of power values into a model trained with machine learning based on log files of data collected from a plurality of power sensors coupled with a plurality of televisions; determine, based on output from the model generated with the array of power values, the television is active; and execute, responsive to the determination that the television is active, an action to evaluate a performance of content rendered by the television.

2. The system of claim 1, comprising: the computing system to receive the time series data from the power sensor.

3. The system of claim 1, comprising the computing system to: receive second time series data from the power sensor; apply a sliding window to construct a second array of power values based on a combination of the second time series data and the time series data; determine, via the model, the television is inactive based on the second array of power values; and prevent execution of the action responsive to the determination that the television is inactive.

4. The system of claim 1, comprising the computing system to: establish a wireless communication session with the power sensor; and receive, via the wireless communication session, the time series data.

5. The system of claim 1, comprising the computing system to: identify an entry in the array of power values that is null; apply an interpolation technique to the array of power values to generate a value to populate the entry; and construct the array of power values with the value generated via the interpolation technique.

6. The system of claim 1, comprising the computing system to: receive, responsive to execution of the action, feedback via an input interface; and determine the performance of the content rendered via the television based on the feedback.

7. The system of claim 1, comprising the computing system to: receive, responsive to execution of the action, feedback via an input interface; and provide, via a network, the feedback to a remote computing system to cause the remote computing system to update the model based on the feedback.

8. The system of claim 1, comprising the computing system to: receive, responsive to execution of the action, feedback via an input interface; and provide, via a network, the feedback to a remote computing system to cause the remote computing system to control subsequent content rendered via the television based on the feedback.

9. The system of claim 1, comprising: the computing system to receive, via a network, the model from a remote computing system, wherein the remote computing system is configured to: receive the log files of data collected from the plurality of power sensors coupled with the plurality of televisions; apply a data cleaning technique to the data of the log files to generate a cleaned data set; apply a normalization technique to the cleaned data set to generate a normalized data set; remove, from the normalized data set, spikes via a filter to generate a filtered data set; interpolate missing values in the filtered data set to generate a training data set; and train, via machine learning, the model with the training data set.

10. The system of claim 1, wherein the model is trained with a training data set generated based on the log files of data collected from the plurality of power sensors, wherein the training data set includes labels generated based on output of a deterministic function applied to the data of the log files.

11. The system of claim 1, wherein the model comprises a long short-term memory network.

12. A method of state-based performance evaluation of televisions via machine learning, comprising: constructing, by a computing system comprising one or more processors coupled with memory, an array of power values from time series data collected by a power sensor coupled with a television; inputting, by the computing system, the array of power values into a model trained with machine learning based on log files of data collected from a plurality of power sensors coupled with a plurality of televisions; determining, by the computing system, based on output from the model generated with the array of power values, the television is active; and executing, by the computing system, responsive to the determination that the television is active, an action to evaluate a performance of content rendered by the television.

13. The method of claim 12, comprising: receiving, by the computing system, the time series data from the power sensor.

14. The method of claim 12, comprising: receiving, by the computing system, a second time series of power values from the power sensor; applying, by the computing system, a sliding window to construct a second array of power values based on a combination of the second time series of power values and the time series data; determining, by the computing system via the model, the television is inactive based on the second array of power values; and preventing, by the computing system, execution of the action responsive to the determination that the television is inactive.

15. The method of claim 12, comprising: establishing, by the computing system, a wireless communication session with the power sensor; and receiving, by the computing system via the wireless communication session, the time series data.

16. The method of claim 12, comprising: identifying an entry in the array of power values that is null; applying, by the computing system, an interpolation technique to the array of power values to generate a value to populate the entry; and constructing, by the computing system, the array of power values with the value generated via the interpolation technique.

17. The method of claim 12, comprising: receiving, by the computing system responsive to execution of the action, feedback via an input interface; and determining, by the computing system, the performance of the content rendered via the television based on the feedback.

18. The method of claim 12, comprising: receiving, by the computing system, responsive to execution of the action, feedback via an input interface; and providing, by the computing system via a network, the feedback to a remote computing system to cause the remote computing system to update the model based on the feedback.

19. The method of claim 12, comprising: receiving, by the computing system, responsive to execution of the action, feedback via an input interface; and providing, by the computing system via a network, the feedback to a remote computing system to cause the remote computing system to control subsequent content rendered via the television based on the feedback.

20. The method of claim 12, comprising: receiving, by the computing system via a network, the model from a remote computing system, wherein the remote computing system is configured to: receive the log files of data collected from the plurality of power sensors coupled with the plurality of televisions; apply a data cleaning technique to the data of the log files to generate a cleaned data set; apply a normalization technique to the cleaned data set to generate a normalized data set; remove, from the normalized data set, spikes via a filter to generate a filtered data set; interpolate missing values in the filtered data set to generate a training data set; and train, via machine learning, the model with the training data set.

04; 06; 04; 04

edspap.20250071376