Open Access. Powered by Scholars. Published by Universities.®
- Publication Year
- File Type
Articles 10861 - 10890 of 11767
Full-Text Articles in Entire DC Network
Classifying Aspect Ratios Of Images, Andrew Gallagher
Classifying Aspect Ratios Of Images, Andrew Gallagher
Defensive Publications Series
Disclosed herein is a mechanism for detecting and/or adjusting aspect ratio in images and videos. For example, the mechanism can include analyzing a digital image having pixels by identifying sets of pixels corresponding to one or more objects in the digital image, analyzing the set of pixels corresponding to at least one object in the image to determine an estimated image aspect ratio, and applying an image transform to the digital image based on the estimated image aspect ratio to generate a modified digital image. In a more particular example, the one or more objects can correspond to faces, animals, …
Machine Learning To Disable Applications From Using Background Resources Except At Appropriate Times, Thomas Price, Eytan Lerba
Machine Learning To Disable Applications From Using Background Resources Except At Appropriate Times, Thomas Price, Eytan Lerba
Defensive Publications Series
Generally, the present disclosure is directed to disabling computing applications from using background resources (e.g., processing or data transmission) except at appropriate times (e.g., times when the user is likely to use the electronic device). In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict an appropriate time for an application to use resources based on device usage data.
Machine Learning To Select Best Network Access Point, Thomas Price, Eytan Lerba
Machine Learning To Select Best Network Access Point, Thomas Price, Eytan Lerba
Defensive Publications Series
Generally, the present disclosure is directed to selecting a best network access point when multiple are available (e.g., including cell towers, WiFi access points, etc.). In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict a best network access point based on device usage data.
Configuring Alarm Setting Using Machine Learning, Thomas Price, Eytan Lerba
Configuring Alarm Setting Using Machine Learning, Thomas Price, Eytan Lerba
Defensive Publications Series
Generally, the present disclosure is directed to configuring the initial time set in an alarm system. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict a time to set an alarm based on historical alarm records and/or other contextual data.
Machine Learning To Select Media Playback Volume, Thomas Price, Eytan Lerba
Machine Learning To Select Media Playback Volume, Thomas Price, Eytan Lerba
Defensive Publications Series
Generally, the present disclosure is directed to setting a media playback volume of an electronic device. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict an appropriate media playback volume based on device usage data.
Machine Learning To Select Input Language On A Software Keyboard, Thomas Price, Eytan Lerba
Machine Learning To Select Input Language On A Software Keyboard, Thomas Price, Eytan Lerba
Defensive Publications Series
Generally, the present disclosure is directed to selecting an input language on a software keyboard. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict an input language for a keyboard based on device usage data.
Machine Learning To Select Screen Brightness Level, Thomas Price, Eytan Lerba
Machine Learning To Select Screen Brightness Level, Thomas Price, Eytan Lerba
Defensive Publications Series
Generally, the present disclosure is directed to setting a brightness of a screen of an electronic device (e.g., during media viewing or media playback). In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict an appropriate screen brightness level based on device usage data.
Machine Learning To Automatically Lock Device Screen At Opportune Time, Thomas Price, Eytan Lerba
Machine Learning To Automatically Lock Device Screen At Opportune Time, Thomas Price, Eytan Lerba
Defensive Publications Series
Generally, the present disclosure is directed to determining when to lock a screen of an electronic device such as, for example, a smartphone. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict an appropriate lock screen behavior based on device usage data.
Generating Travel Itineraries Based On Travel History Of Similar Users, Alexandre Chinien
Generating Travel Itineraries Based On Travel History Of Similar Users, Alexandre Chinien
Defensive Publications Series
Generally, the present disclosure is directed to generating a travel itinerary for a user based on travel history of similar users. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to generate a travel itinerary for a user based on travel history data from one or more users.
Machine Learning Approach For User Interface Testing, Hassan Abolhassani
Machine Learning Approach For User Interface Testing, Hassan Abolhassani
Defensive Publications Series
A system and method are disclosed that enable automated testing of a user interface. The testing system includes a machine learning (ML) algorithm to test a user interface. The ML approach includes collecting training samples, creating an ML model and using the model to test the user interface. Training samples are collected by providing the users with diff files with the differences highlighted. A user is provided with options to specify if the differences are acceptable or not. The user classification and other attributes are used to train the model. When a new diff is created it may be fed …
Advertising Mechanism For Predictive Consumption Of Content At A Variable Entry Point, Morgan Dollard
Advertising Mechanism For Predictive Consumption Of Content At A Variable Entry Point, Morgan Dollard
Defensive Publications Series
A mechanism that quantifies future traffic associated with respective segments of content items. A content item may be received from a content provider and segments of the received content item may be identified. Subsequently, the identified segments may be classified based on the type of content present in the content item during the segments. The future consumption of the segments of the content item may be predicted based on the type of content present and the segments may be ranked and priced accordingly. The content item may then be presented including advertisements during the segments of the content item.
User Assistant Device, Cristian Cibils Bernardes
User Assistant Device, Cristian Cibils Bernardes
Defensive Publications Series
The present disclosure provides a device that enables a user to automatically trigger an action to be performed by one or more computing devices with the push of a button. In particular, the device can be a wireless button coupled to a user assistant enabled computing device. Existing user assistant enabled computing devices can be configured to accept user text or voice commands or queries to perform certain actions. To repeat a common action the user must repeatedly enter the same text or voice commands or queries. Despite the frequency of a particular command or query, user assistant enabled devices …
Preferential Resource Delivery Via Web Proxy Or Web Browser, Dean Kenneth Jackson, Daniel Klein
Preferential Resource Delivery Via Web Proxy Or Web Browser, Dean Kenneth Jackson, Daniel Klein
Defensive Publications Series
The subject disclosure relates to prioritizing the loading of files referenced by the HyperText Markup Language (HTML) of a webpage, so that the time required to load the webpage is reduced, for example, as perceived by an end user. The HTML for a webpage can be parsed to identify what types of resources are referenced (e.g., a cascading style sheets (CSS) file, image files, JavaScript), and the bandwidth associated with downloading the resources can be assessed. Based on requested resources and bandwidth, the system can prioritize the loading of resources, in order to load portions of the webpage quickly so …
Handwriting Recognition For Ides With Unicode Support, Michal Luszczyk, Sandro Feuz
Handwriting Recognition For Ides With Unicode Support, Michal Luszczyk, Sandro Feuz
Defensive Publications Series
This disclosure describes techniques for handwriting recognition for Unicode text entered via touch or other input. For example, Unicode text may be part of program source code in some programming languages. The handwritten code is provided as an input to a trained machine learning model. The model produces possible matches for the user entered text along with a score for each match. The user, e.g., a programmer, can select the appropriate match which is then added to the source code. The model is trained with sample handwritten text for known programs, and can also utilize additional factors such as program …
Generation Of Meeting Transcript Using Speech Recognition, Eric Bellamy
Generation Of Meeting Transcript Using Speech Recognition, Eric Bellamy
Defensive Publications Series
Participants in meetings conducted via audio or videoconferencing often engage in notetaking. This can cause participants to be less engaged. This disclosure describes techniques to automatically generate a meeting transcript. At the beginning of a meeting, participant identities and voices are learned as participants introduce themselves. During the meeting, participants can enable transcript generation with voice commands. Speech-to-text techniques are used to generate a text transcript of the meeting that associates speech with corresponding participants.
Hyper Suggestion Spans For Improved Text Correction, Glen Shires, Jakob Foerster, Diego Melendo Casado
Hyper Suggestion Spans For Improved Text Correction, Glen Shires, Jakob Foerster, Diego Melendo Casado
Defensive Publications Series
Improved techniques for correcting text transcriptions of voice input is described herein. A computing device may perform speech recognition to transcribe voice input (e.g., speech) into text, enable a user to select any of one or more consecutive words in the text transcription, and, in response, determine and output a set of alternative hypotheses for the selected one or more consecutive words, so that the user may select one alternative hypothesis in the set of alternative hypotheses in order to replace the selected one or more consecutive words with the alternative hypothesis.
Minimal Restriping For Data Center Expansion, Shizhen Zhao
Minimal Restriping For Data Center Expansion, Shizhen Zhao
Defensive Publications Series
A system and a method are proposed to minimize restriping during data center expansion. Minimal restriping technique takes into account the old data center topology, and aims to minimize restriping between the new topology and the old topology. A low complexity minimal restriping algorithm proposed is used to minimize restriping. The topology solution is also guaranteed to have high network capacity, because it satisfies superblock ports, spineblock ports,and superblock-spineblock level balancedness constraints. The constraints may consider the number of links in original topology and the algorithm may determine the optimum new topology that minimizes restriping. The disclosed technique may not …
Virtual Reality Re-Centering, Chris Mckenzie
Virtual Reality Re-Centering, Chris Mckenzie
Defensive Publications Series
A virtual reality player often views a virtual reality scene of a virtual world while the player is located in a physical world via a head-mounted display. If the physical world is a virtual reality room, the scene can be re-centered within a known region, where stationary objects in the physical world (furniture, walls, etc.) may be accounted for via data such as maps or dimensions of the virtual reality room. However, in instances where the physical world is not a virtual reality room, such as a random standing space or a couch of a player’s house, a virtual reality …
Adaptive Screen Layouts Based On Viewer Proximity From Home Devices, Daniele Midi, William Schilit
Adaptive Screen Layouts Based On Viewer Proximity From Home Devices, Daniele Midi, William Schilit
Defensive Publications Series
Smart devices in the form of personal assistants are becoming more common and more dynamic. In addition to audibly exchanging information with a user via a mechanism such as a smart speaker, a personal assistant may now include a display to present useful ambient information to a user, such as time, weather, calendar information, or upcoming events. Techniques are described for adapting a screen layout of such a display based on proximity of a viewer relative to the display, allowing an effective and efficient presentation of the useful, ambient information.
Dynamic Slider Scaling, Michael Alger
Dynamic Slider Scaling, Michael Alger
Defensive Publications Series
Techniques are described that enable dynamic slider scaling. In a VR user interface with dynamic slider scaling, the slider control expands when the user selects the control. The expanded control offers better control to adjust the slider and may include additional information that is not available with conventional slider controls. When the user has completed the adjustment and releases the slider control, it reverts to its original size and shape to conserve space in the VR user interface.
Dynamic Notification Settings For New Or Unused Notification Types, Joseph Cohen, Justin Lewis
Dynamic Notification Settings For New Or Unused Notification Types, Joseph Cohen, Justin Lewis
Defensive Publications Series
A mechanism to provide notification settings to a user may be based on the interaction of the user at a content item service. For example, a likelihood of the user receiving particular notifications may be identified. The notifications may identify a particular content item at the content item service or a feature of the content item service. Settings for such notifications that have not been previously specified by the user may be identified. Subsequently, the notification settings may be provided to the user so that the user may enable or disable the receiving of future notifications that may be more …
Dynamic Notification Settings For New Or Unused Notification Types, Joseph Cohen, Justin Lewis
Dynamic Notification Settings For New Or Unused Notification Types, Joseph Cohen, Justin Lewis
Defensive Publications Series
A mechanism for dynamically populating a user setting menu for notification settings is presented herein. The mechanism may present notification settings for new notification types and/or notification sources from which a user has not received a notification. The notification settings may be dynamically added to a user settings menu based on a user affinity score to a notification source and/or whether the notification type has been received by the user. Accordingly, the mechanism enables displaying notification settings to the user that the user is likely to receive in the future prior to the user actually receiving the notification. This may …
Smart-Voice Invocation Of Scenes In Home-Automation Systems, Daniele Midi, William Schilit, Tianchun Yang
Smart-Voice Invocation Of Scenes In Home-Automation Systems, Daniele Midi, William Schilit, Tianchun Yang
Defensive Publications Series
Home environments are rapidly becoming more and more automated via the use of voice-recognizing personal assistants. Home-automation systems using a voice-recognizing personal assistant are capable of configuring smart devices of a home, such as lighting, security systems, or environmental controls, to particular operational settings which, when aggregated across multiple smart devices, constitute a particular scene. Techniques for a home-automation system to configure, or “set”, a scene using smart-voice invocation are described.
Domain Restricted Secure Podcasts, Dustin Beutin
Domain Restricted Secure Podcasts, Dustin Beutin
Defensive Publications Series
This disclosure describes techniques to securely provide access to audio files across desktop and mobile devices. The audio files, e.g., audio recordings that are part of a podcast, are converted to multimedia files of similar filesize. The converted files, which can optionally include a graphic related to the audio, are uploaded to secure, view-only cloud storage and configured for access by selected users, e.g., users within a particular domain. Access to the files is provided by a widget that can enable users to play the audio via any desktop or mobile device that can access the cloud storage service.
Proximity-Based Single-Tap Meeting Room Selection, Xavier Lepaul, Josefin Karlsson, Carolina Sartorius, Megan Boundey, Alex Usbergo
Proximity-Based Single-Tap Meeting Room Selection, Xavier Lepaul, Josefin Karlsson, Carolina Sartorius, Megan Boundey, Alex Usbergo
Defensive Publications Series
A meeting participant can join a scheduled or ad-hoc meeting from a conference room by manual entry of a meeting code into meeting room audio or video conferencing equipment. This is a tiresome and error-prone process that requires multiple steps. This disclosure describes techniques that enables users to join a meeting at a nearby conference room using a simple user interface. With user permission, a mobile device of the user determines user proximity to a conference rooms based on signals transmitted by conference rooms. The identity of the available conference room(s) is used to generate and present a user interface …
Machine Learning For Hardware Simulation, N/A
Machine Learning For Hardware Simulation, N/A
Defensive Publications Series
Generally, the present disclosure is directed to improving hardware simulation through machine learning. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict hardware test results based on data extracted from hardware.
Injection Molding Eyepieces For Head Mounted Display, Hongqin Shi, Yeh-Jiun Tung
Injection Molding Eyepieces For Head Mounted Display, Hongqin Shi, Yeh-Jiun Tung
Defensive Publications Series
A technique is provided for manufacturing eyepieces for a head mounted display by forming the eyepiece in two parts: a coated, smaller insert part with a first coating (e.g. chiolite), and a second, larger supporting part with an opening or gap for the coated smaller part. The inset and supporting parts are formed from an eyepiece material via injection molding, and the inset part is then coated with the first coating. The inset is inserted into the gap of the supporting part, and the two parts are cast with an index matching material. The casting forms a front and back …
Contextual Advertising Based On Content Recognition In A Video, Jason Mayes
Contextual Advertising Based On Content Recognition In A Video, Jason Mayes
Defensive Publications Series
Generally, the present disclosure is directed to providing relevant advertisements based on the visual content of a video. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to determine a relevant advertisement and/or a relevant time for the relevant advertisement based on image data taken from a video.
Identifying And Categorizing Utility Power Anomalies, N/A
Identifying And Categorizing Utility Power Anomalies, N/A
Defensive Publications Series
Generally, the present disclosure is directed to predicting utility power anomalies such as, for example, outage events. In particular, in some implementations, the systems and methods of the present disclosure can include or otherwise leverage one or more machine-learned models to predict utility power anomalies based on historical and/or current utility power data.
Consistent Hashing For Generating Protobuf Field Numbers, Tomasz Madejski, Rob Shakir
Consistent Hashing For Generating Protobuf Field Numbers, Tomasz Madejski, Rob Shakir
Defensive Publications Series
To convert a YANG-based model into a protobuf format, a unique field identification numbers needs to be assigned to each node in the YANG schema tree. Field identification numbers can be generated by calculating a numeric hash of characteristics of the YANG schema node; for example, its identifier or schema tree path. The resulting hash values provide backwards compatible field identification numbers suitable for protobuf field numbers.