In some embodiments, the private network may include at least one first device that captures information about its surrounding environment, such as data about the people and/or objects in the environment. The first device may receive a set of potential content sent from a server external to the private network. The first device may select at least one piece of content to present from the set of potential content based in part on the people/object data and/or a score assigned by the server to each piece of content. The private network may also include at least one second device that receives the captured people/object data sent from the first device. The second device may also receive a set of potential content sent from the server external to the private network. The second device may select at least one piece of content to present from the set of potential content based in part on the people/object data sent from the first device and/or a score assigned by the server to each piece of content. Using the private network to communicate the people/object data between devices may preserve the privacy of the user since the data is not sent to the external server. Further, using the obtained people/object data to select content enables more personalized content to be chosen.

[…]

 

  • urther, although not shown in this particular way, in some embodiments, the client device 134 may collect people/object data 136 using one or more sensors, as discussed above. Also, as previously discussed, the raw people/object data 136 may be processed by the sensing device 138, the client device 134, and/or a processing device 140 depending on the implementation. The people/object data 136 may include the data described above regarding FIG. 7 that may aid in recognizing objects, people, and/or patterns, as well as determining user preferences, mood, and so forth.
  • [0144]
    After the client device 134 is in possession of the people/object data 136, the client device 134 may use the classifier 144 to score each piece of content 132. In some embodiments, the classifier 144 may combine at least the people/object data 136, the scores provided by the server 67 for the content 132, or both, to determine a final score for each piece of content 132 (process block 216), which will be discussed in more detail below.
  • [0145]
    The client device 134 may select at least one piece of content 132 to display based on the scores (process block 218). That is, the client device 134 may select the content 132 with the highest score as determined by the classifier 144 to display. However, in some embodiments, where none of the content 132 generate a score above a threshold amount, no content 132 may be selected. In those embodiments, the client device 134 may not present any content 132. However, when at least one item of content 132 scores above the threshold amount and is selected, then the client device 134 may communicate the selected content 132 to a user of the client device 134 (process block 220) and track user interaction with the content 132 (process block 222). It should be noted that when more than one item of content 132 score above the threshold amount, then the item of content 132 with the highest score may be selected. The client device 134 may use the tracked user interaction and conversions to continuously train the classifier 144 to ensure that the classifier 144 stays up to date with the latest user preferences.
  • [0146]
    It should be noted that, in some embodiments, the processing device 140 may receive the content 132 from the server 67 instead of, or in addition to, the client device 134. In embodiments where the processing device 140 receives the content 132, the processing device 140 may perform the classification of the content 132 using a classifier 144 similar to the client device 134 and the processing device 140 may select the content 132 with the highest score as determined by the classifier 144. Once selected, the processing device 140 may send the selected content 132 to the client device 134, which may communicate the selected content 132 to a user.
  • […]
  • The process 230 may include training one or more models of the classifier 144 with people/object data 136, locale 146, demographics 148, search history 150, scores from the server 67, labels 145, and so forth. As previously discussed, the classifier 144 may include a support vector machine (SVM) that uses supervised learning models to classify the content 132 into one of two groups (e.g., binary classification) based on recognized patterns using the people/object data 136, locale 146, demographics 148, search history 150, scores from the server 67, and the labels 145 for the two groups of “show” or “don’t show.”

 

Source: US20160260135A1 – Privacy-aware personalized content for the smart home – Google Patents

 

They have thought up around 140 ways that this can be used…