Microsoft Technology Licensing, LLC

Yong Rui

Anoop Gupta

Ross Garrett Cutler

Yunqiang Chen

Cha Zhang

Dinei A Florencio

Andrew David Wilson

Desney S Tan

Daniel Morris

Nikunj Raghuvanshi

Jeannette M Wing

Li-wei He

Li-Wei He

Alejandro Acero

Hong-Jiang Zhang

Michael F Cohen

Xian-Sheng Hua

Paul A Viola

Tao Mei

Xinding Sun

Francis C Li

Pei Yin

Guo-Jun Qi

Bin Yu

Johnathan J Cadiz

Shipeng Li

Henrique Sarmento Malvar

Zicheng Liu

Ivan Jelev Tashev

Andres Vega-Garcia

Kuiyuan Yang

Lei Zhang

Wei-Ying Ma

Xuedong D Huang

Eric H Rudolph

Zhiwei Li

Sasa Junuzovic

Rui Cai

Hongjiang Zhang

Di Li

Jiawei Gu

Roy A Leban

Jonathan Thomas Grudin

Pengfei Xu

Jianlong Fu

Yan-Feng Sun

Prasun Dewan

Ting Yao

Chunyan Liu

Xuedong David Huang

Chun-Che Wu

Baoxun Wang

Ting Yu

Samuel Y Shen

Shinichi Manaka

Michael Wallick

Le-wei He

Qiong Liu

An image chat application generates comments to images based on features of the images. In one example, the image chat application searches through a repository of stored image-comment pairs to identify a stored image that is similar to the image, and generates a comment to the image based on an identified stored image-comment pair. In another example, the image chat application may identify and tag particular objects that dominate an image, and may generate a comment to the image based on characteristics of those particular objects. In this second example, the image chat application further generates a comment to the image based on comments previously associated with the identified tag.

Identifying image comments from similar images

Techniques and constructs to facilitate suggestion of image-based search queries can provide personalized trending image search queries. The constructs may enable identification of trending image searches and further personalize those trending image search queries for an identified user based on information about on the user's search history and the search histories of other users. The constructs also may select a representative image for display to the user, such that selection of the representative image will execute the search query. The representative image may be selected from a plurality of candidate images based on its burstiness.

Personalized trending image search suggestion

Computer-readable media, computer systems, and computing devices for initiating a search function, such as presentation of a search box or initiation of a search, is provided. In one embodiment, the method includes detecting movement of a selector from within a display area to an edge of the display area. Such a selector can be controlled by an input device coupled to a user device. In response to detecting movement of the selector from within the display area to the edge of the display area, a search-query input area associated with a search engine is presented within a display screen view.

Gesture-based search

A privacy preserving sensor apparatus is described herein. The privacy preserving sensor apparatus includes a microphone that is configured to output a signal that is indicative of audio in an environment. The privacy preserving sensor apparatus further includes feature extraction circuitry integrated in the apparatus with the microphone, the feature extraction circuitry configured to extract features from the signal output by the microphone that are usable to detect occurrence of an event in the environment, wherein the signal output by the microphone is unable to be reconstructed based solely upon the features.

Privacy preserving sensor apparatus

Technologies pertaining to calibration of filters of an audio system are described herein. A mobile computing device is configured to compute values for respective filters, such as equalizer filters, and transmit the values to a receiver device in the audio system. The receiver device causes audio to be emitted from a speaker based upon the values for the filters.

Dynamic calibration of an audio system

Video description generation using neural network training based on relevance and coherence is described. In some examples, long short-term memory with visual-semantic embedding (LSTM-E) can maximize the probability of generating the next word given previous words and visual content and can create a visual-semantic embedding space for enforcing the relationship between the semantics of an entire sentence and visual content. LSTM-E can include a 2-D and/or 3-D deep convolutional neural networks for learning powerful video representation, a deep recurrent neural network for generating sentences, and a joint embedding model for exploring the relationships between visual content and sentence semantics.

Jointly modeling embedding and translation to bridge video and language

Techniques and constructs to facilitate automatic tagging can provide improvements in image storage and searching. The constructs may enable training a deep network using tagged source images and target images. The constructs may also train a top layer of the deep network using a personal photo ontology. The constructs also may select one or more concepts from the ontology for tagging personal digital images.

Tagging personal photos with deep networks

Technologies pertaining to provision of customized audio to each listener in a plurality of listeners are described herein. A sensor outputs data that is indicative of locations of multiple listeners in an environment. The data is processed to determine locations and orientations of the respective heads of the multiple listener in the environment. Based on the locations and orientations of heads of the listeners in the environment, for each listener, respective customized audio signals are generated. The customized audio signals are transmitted to respective beamforming transducers. The beamforming transducers directionally output customized beams for the first listener and the second listener based upon the customized audio signals and locations of the heads of the listeners.

Enhanced spatial impression for home audio

A method of acquiring a set of images useable to 3D model a physical object includes imaging the physical object with a camera, and displaying with the camera a current view of the physical object as imaged by the camera from a current perspective. The method further includes displaying with the camera a visual cue overlaying the current view and indicating perspectives from which the physical object is to be imaged to acquire the set of images.

User interface for three-dimensional modeling

Estimation of available bandwidth on a network uses packet pairs and spatially filtering. Packet pairs are transmitted over the network. The dispersion of the packet pairs is used to generate samples of the available bandwidth, which are then classified into bins to generate a histogram. The bins can have uniform bin widths, and the histogram data can be aged so that older samples are given less weight in the estimation. The histogram data is then spatially filtered. Kernel density algorithms can be used to spatially filter the histogram data. The network available bandwidth is estimated using the spatially filtered histogram data. Alternatively, the spatially filtered histogram data can be temporally filtered before the available bandwidth is estimated.

Methods and systems for estimating network available bandwidth using packet pairs and spatial filtering

Systems and methods for detecting people or speakers in an automated fashion are disclosed. A pool of features including more than one type of input (like audio input and video input) may be identified and used with a learning algorithm to generate a classifier that identifies people or speakers. The resulting classifier may be evaluated to detect people or speakers.

Identification of people using multiple types of input

Kernelized spatial-contextual image classification is disclosed. One embodiment comprises generating a first spatial-contextual model to represent a first image, the first spatial-contextual model having a plurality of interconnected nodes arranged in a first pattern of connections with each node connected to at least one other node, generating a second spatial-contextual model to represent a second image using the first pattern of connections, and estimating the distance between corresponding nodes in the first spatial-contextual model and the second spatial-contextual model based on a relationship with adjacent connected nodes to determine a distance between the first image and the second image.

Kernelized spatial-contextual image classification

Growing community of inventors

Sammamish, WA, United States of America

Yong Rui