Digital Commons Network^™

Secure Audio Processing, Thad Hughes, Ignacio Lopez Moreno, Aleksandar Kracun, Pedro Moreno

Defensive Publications Series

Automatic speech recognizers (ASR) are now nearly ubiquitous, finding application in smart assistants, smartphones, smart speakers, and other devices. An attack on an ASR that triggers such a device into carrying out false instructions can lead to severe consequences. Typically, speech recognition is performed using machine learning models, e.g., neural networks, whose intermediate outputs are not always fully concealed. Exposing such intermediate outputs makes the crafting of malicious input audio easier. This disclosure describes techniques that thwart attacks on speech recognition systems by moving model inference processing to a secure computing enclave. The memory of the secure enclave and signals …

Directed Messages With Natural Identifiers, Oliver Gaymond, Lidia Gaymond

Defensive Publications Series

The present disclosure describes systems and methods that provide for routing messages among devices located around a physical location without requiring the sender to know the current location of the recipient. More particularly, the present disclosure provides for combining a plurality of devices, such as smart speakers, home assistants, and/or the like, located around a physical location, such as an office, building, and/or the like, with intelligent voice matching capabilities to enable inter-room messaging by addressing the individual recipient without needing to know the room they are currently located in. In some implementations, the devices can monitor their surroundings for …

Automatically Emphasizing Displayed Content In Video Communication Interfaces, Bailiang Zhou

Defensive Publications Series

Video calls, video conferences, and similar real-time communications between devices display users and other subject matter to other users participating in the communication in a user interface. This disclosure describes techniques in which display of particular video content can be emphasized in the user interface based on particular characteristics of the video content. The emphasized display can be used to automatically highlight video content at appropriate times, allow easier viewing of video content by other users in a communication or other video capture or display application, etc.

Voice Input For Authentication, Nick Felker, Shuyang Chen, Sachit Mishra

Defensive Publications Series

Recent years have seen a proliferation of systems designed to respond to voice, e.g., smart speakers, smartphones, and other voice-activated devices. Such systems are convenient and intuitive to use. However, they also open the possibility for malicious actors to gain control of sensitive information through fraudulent use of speech commands. For example, a malicious actor can spoof the voice recognition system through a recording of the user’s voice or by using a computer to synthesize the speech mimicking the user. The techniques of this disclosure thwart such attacks by providing a random phrase for the user to repeat and verifying …

Hands-Free Voice Assistant Application In A Vehicle, Ashwin Gururaghavendran

Defensive Publications Series

This disclosure describes the integration of a digital voice assistant application with a vehicular hands-free (e.g., Bluetooth) calling system. A user pairs their mobile device, e.g., smartphone, tablet, etc., with a vehicular hands-free system to facilitate hands-free calling. The dialing of a designated number using the mobile device serves as an indicator to the Bluetooth stack and/or the operating system to trigger the voice assistant application. When triggered, the voice assistant application receives commands and input via voice over Bluetooth. Responses from the voice assistant application are played over the speaker system.

Adaptive Hvac Operation To Reduce Disruptive Fan Noise Levels During Noise-Sensitive Events, Paul Gyugyi

Defensive Publications Series

Some heating, ventilation, and air condition (HVAC) systems produce a significant amount of noise. Such systems can have a fan that is active when heating or air conditioning is occurring to circulate air. Noise generated by such a fan can be disruptive, especially if one or more nearby persons are participating in a noise-sensitive activity, such as watching television or talking. Embodiments are focused on determining when such noise-sensitive events are occurring or are likely to occur and adjusting control of an HVAC system to decrease or eliminate the sound or noise generated by the HVAC system while the noise-sensitive …

Blockchain Based Rating And Review Platform, Dong Lin

Defensive Publications Series

This disclosure describes a blockchain based rating and review platform for web content. The platform provides a distributed and open access framework for ratings and reviews. With user permission, data corresponding to the rating, review, evaluation, etc., of an item (article, picture, sound clip, etc.) of web content by a user is stored in a blockchain record. A single blockchain record can contain one or more review actions. The blockchain record serves as a trusted ledger of reviews and ensures that the rating records are immutable. Rating data retrieved from the platform can be utilized for analytics and to assist …

Automatic Evaluation Of Contractual Terms, Alex Zheng

Defensive Publications Series

This disclosure describes use of machine learning and natural language processing techniques to assist users in reviewing terms and conditions (T&C), e.g., provided by online product and service providers. A database of T&C is created and indexed according to factors such as country, industry, application, etc. The database is utilized to identify T&C content that is common for a given industry application in a specific jurisdiction. When a user provides permission to analyze and provide assistance for a T&C document, such document is analyzed to determine a risk score. Further, any anomalous content in the T&C is flagged. The user …

Automatic Translation And Transcription Of Voicemail, Jan Jedrzejowicz, Alex Zheng

Defensive Publications Series

This disclosure describes automatic translation and transcription of foreign language voicemails into a language of a user’s choice. Language detection, translation, and transcription techniques are employed to generate the text transcript in the preferred language. The user is provided a text transcript of voicemail in their preferred language with an indication that the voicemail was translated.

Concurrent Speculative Execution Of Multiple Communication Applications, Steffen Meschkat, Sandeep Siddhartha

Defensive Publications Series

Upon detecting user context to communicate with a contact, an operating system speculatively launches communication applications that serve as communication channels between the user and the contact. With user permission, the availability of the contact on each of the communication channels is displayed along with the history of the user’s communication with the contact. The user interface enables selection of a suitable communication channel for communicating with the contact without having to switch applications.

Reward Calculation Based On Quality Of Task Completion, Luis Benavides Oyaga, Ravi Solanki

Defensive Publications Series

This disclosure describes techniques to incentivize higher quality in tasks performed by users as part of goal based advertising. Machine learning techniques are utilized to determine a quality score for tasks completed by users. Tasks are assigned in parallel to multiple users based on the quality score. The quality score is used to determine and provide rewards upon task completion.

Goal Completion Based Content Access, Luis Benavides Oyaga, Ravi Solanki

Defensive Publications Series

This disclosure describes techniques to use goal based advertising. Users are enabled to unlock ad-free content or obtain services in exchange for completion of advertiser specified actions. Advertisers and other parties define the actions as predefined goals. The available goals from various parties are ranked, e.g., using an auction. The ranking can be based on the probability with which a user is likely to complete actions corresponding to each goal. The number and type of actions to be offered to a user are determined. When users permit use of user data, the offers can be tailored to user interests. The …

Secure Modular Scalable Management For Rack Mount Switches, John Norton, David Koenen

Defensive Publications Series

Next generation rack mount switches will be required to support a higher bandwidth densities through the faceplate as switch ASICs grow in lane count. Management interfaces in switches currently take up that valuable faceplate space. In addition, the server market is looking for an ever increasing ability to scale out infrastructure and configure to specific workloads. This scalable management interface for switches is external to the switch chassis, integrated into the rack, and physically secure.

Smart Headgear For Display Of Transcribed Or Translated Speech, Sara Basson, Dimitri Kanevsky

Defensive Publications Series

Natural language translation and transcription technologies can now enable x real-time conversation between speakers of mutually unintelligible languages. Such technologies also assist hearing-impaired speakers by providing real-time transcription. Currently, the transcribed or translated speech is typically displayed on a screen that is away from the speaker, e.g., on a hand-held device. Such type of display is not conducive to maintaining the eye-to-eye contact necessary for natural conversation. The techniques of this disclosure describe smart headgear that integrates a display capable of streaming transcribed or translated speech to a listener. The smart headgear integrates or interoperates with natural language processing hardware, …

Disambiguation Of Entity References In Natural Language Interpretation, Steffen Meschkat

Defensive Publications Series

This disclosure describes techniques for disambiguation of entity references in user input. The techniques can be implemented in computing systems that store data as identifiable entities along with relationships between the entities and are effective in user input with possible ambiguities. User input is scanned for references to entities generically. If an input name refers to more than one known entity known, and another entity name is detected in the same user input, the relationships between the entities are used to disambiguate the input.

Concurrent Speculative Execution Of Selector Applications, Steffen Meschkat, Ushasree Kode, Christopher Lane

Defensive Publications Series

Upon detecting a user context to select content items, an operating system speculatively launches software applications that provide the particular type of content items. A user interface is provided that enables a user to select items from collections of items provided by the applications. The user interface enables a user to perform such selection without having to launch additional applications or switch between applications.

Disambiguation Of Entity References Using Related Entities, Steffen Meschkat

Defensive Publications Series

This disclosure describes techniques for disambiguation of entity references in natural language processing. The disambiguation of ambiguous entity references is performed by seeking user confirmation of the intended reference. With user permission, the confirmation query or prompt utilizes phrases that include the name of the ambiguous entity and name(s) of related entities. Entities are stored along with names of related entities and designation of the nature of the relationship. The nature of relationship between stored entities is used to construct appropriate conjunctions in disambiguation phrases. The described techniques can be used in any natural language interface, e.g., text or voice, …

Automatic Contextual Recognition Of Hand-Drawn Content, Kyle Williams, Milton Ribeiro Filho, Megan Renshaw

Defensive Publications Series

Digital whiteboards and touchscreens enable users to express concepts using free-hand drawing and writing. Although applications exist that recognize handwritten matter or geometric shapes, such recognition is not based on the user context. This disclosure describes techniques to transform user input on a digital canvas, e.g., hand-drawn strokes on a touchscreen or digital whiteboard, into software editable objects, based on determined context. The context is determined based on user permission for analysis of the strokes. The software editable objects are provided in a format suitable for a software application matched to the context, e.g., as editable graphical structures such as …

Wireless Device For Casting Content To Available Screens, Eric Bellamy

Defensive Publications Series

This disclosure describes a device designed to allow content from a first screen to be wirelessly displayed on a second screen. The device is a wireless mouse-like device that communicates with a monitor or screen. The device is equipped with a first button to select content and to turn on screens by communicating with an infrared receiver in the monitor. A second button on the device is configured to select and move content from one screen to another. The device allows for efficient and quick viewing of content on a larger screen and enables any available monitor to act as …

Cloudslurm: A Multi-Provider Approach For Hpc In The Cloud, Glaucimar Da Silva Aguiar, Andrey Elision Monteiro Brito, Flávio Fonseca, Luís Eduardo Tenório Silva, Lucas Lima Vieira

Defensive Publications Series

This disclosure relates HPC clusters elasticity using cloud nodes. A framework is disclosed that implements an integration of HPC clusters based on the Slurm workload manager and a cloud infrastructure. The main goal is to be transparent for the Slurm user, who may not be familiar with cloud technologies or with a provider-specific API.

Action Prediction From Conversations, Victor Carbune, Alexandru Damian

Defensive Publications Series

This disclosure describes techniques to automatically extract actions and reminders from conversations between users, when users permit analysis of conversations. With user permission, speech recognition, speaker disambiguation, and action extraction, e.g., implemented using trained machine learning models, are utilized to analyze conversations, detect when the conversations include user actions, and determine the owner of the actions. Actions are grouped by user and/or by type. A user interface displays follow-up actions identified from a conversation for review by users and enables users to make corrections.

Hybrid Lifi And Wifi Communication System, Xiangyu Kong, Haijiang Ma

Defensive Publications Series

A system and method are proposed that may transfer information between devices using both LiFi and the WiFi technologies. The hybrid system may use LiFi for downlink traffic and WiFi for uplink traffic. The system may include a LiFi receiver/transmitter and also a WiFi receiver/transmitter in the devices. When a device tries to communicate with a network, the data is uploaded from the device to the router connected to the network through WiFi. The device receives response from the network via the router through LiFi. The disclosed hybrid system incorporates the high download speed of LiFi (up to 4 Gbps), …

Safe Parking: Further Development Epb +, Daniel Hoppe

Defensive Publications Series

Safe Parking: Further Development EPB +

Location-Based Jammer For Supl Terminal, Thippeswamy J. H. M.

Defensive Publications Series

A technique is disclosed that provides a location-based jammer for SUPL (Service User Plane Location) enabled terminals ("SET"), using MAP Supplementary Services. Triggers that activate and deactivate call barring are made when a device enters and leaves a call-restricted area respectively.

Automatic Determination Of Delays And Appointment Rescheduling, Jorim Jaggi, Yannick Stucki, Sandro Feuz, Adrian Roos

Defensive Publications Series

This disclosure describes techniques to automatically predict that a user is likely to be delayed for an upcoming calendar appointment. A trained on-device machine learning model is used to predict delays based on permitted contextual information. Schedule rearrangements are automatically generated, e.g., to rearrange subsequent calendar meetings that are predicted to be affected due to the delay. Rearrangements are presented to the user, and the calendar is updated per user input and preferences. A calendar application or service that implements these techniques can reduce the burden on users to manually reschedule appointments.

Transfer Of Payment Credentials Between Devices, Roman Kalukiewicz, Justin Brickell

Defensive Publications Series

Payment credentials for use in mobile wallets are stored on a user’s mobile device through a provisioning process. When a user switches devices, the new mobile device has to be provisioned with the user’s payment credentials to enable mobile wallet functionality on the new device. Such provisioning can be cumbersome. This disclosure describes secure transfer of payment credentials between mobile devices by physically tapping the devices together. A new mobile device is configured as a payment terminal. The user is instructed to tap the old mobile device against the new mobile device. Upon tapping, the old mobile device transfers payment …

Adaptive Audio Processing Based On Scene Detection, Leonardo Kusumo, Moonseok Kim

Defensive Publications Series

This disclosure describes techniques to automatically adjust audio processing parameters based on acoustic scene detection. A parameter adjustment module utilizes default dynamic range compression settings during initial audio recording using a mobile device. A trained machine learning model is utilized to detect the acoustic scene, based on initial audio signals. Features such as the average loudness, maximum loudness, minimum loudness, acoustic scene, wind noise, and other user context features, as permitted by the user, are provided to the trained model. The model is utilized to adapt audio processing parameters such as the compression profile to the detected scene to improve …

Detecting Rack U Location Using Light Based Messages And Camera, Daniel Pol

Defensive Publications Series

As the amount of data generated daily increases exponentially, the hardware needs to process and store all that information increases also, and the speed with which new hardware can be deployed to service those needs becomes more relevant every day. Most of the automated provisioning solutions today still require a lot of manual steps, especially related to initial hardware inventory and setting up basic network communications.

This disclosure relates to the field of Data Center Infrastructure.

A method is disclosed for using light based messages and cameras to detect device physical location within a data center rack cabinet.

Interactive Web Browser Learning Tool, Eric Bellamy

Defensive Publications Series

Students that suffer from learning disabilities often need specialized educational techniques to enhance learning abilities. Some of these include direct instruction, multi-sensory approaches, and repetition. This disclosure describes techniques to provide an interactive web tool that, with user permission, learns what a student likes and then uses those learned inputs to generate an interactive learning exercise tailored to a student’s learning progress.

Simultaneous Multimodal User Interface, Sandro Feuz, Thomas Deselaers

Defensive Publications Series

This disclosure describes techniques to use a combination of voice based input with other input mechanism in human-computer interfaces. The techniques enable users to utilize in real-time, the suitable input mode(s) in a given context, without having to switch between input modes. With user permission, speech analysis techniques are utilized to analyze user speech and detect when speech includes user instructions, and to determine corresponding actions to be performed. By enabling simultaneous user input via multiple modes, the techniques facilitate effective navigation of complex tasks performed using a computing device.