Amazon Technologies, Inc.

Rawles LLC

Ning Yao

Samuel Henry Chang

Sowmya Gopalan

Qiang Liu

Edward Dietz Crump

Rahul Agrawal

Weijian Wang

William Spencer Worley, Iii

Yue Liu

Kavitha Velusamy

Vijay Kamarshi

Manika Puri

Olusanya Temitope Soyannwo

Renwei Yan

Yuhua Zhu

Charles Shearer Dorner

Wai Chung Chu

Menashe Haskin

Eva Manolis

Douglas J Gradt

Robert Warren Sjoberg

Robert Ramsey Flenniken

Kevin Wayne Arthur

Jongwoo Lee

Alexander Adrian Hugh Davidson

Maggie McDowell

Helder Jesus Ramalho

Paul Matthew Bombach

Techniques are described for detecting a periphery of a surface based on a point set representing the surface. The surface may correspond to a display medium upon which content is projected. A shape model may be matched and aligned to a contour of the point set. A periphery or edge of the surface and corresponding display medium may be determined based on the aligned shape model.

Shape-based edge detection

Shape-Based Edge Detection

A number of images of an environment may be obtained over time by a number of sensors. The images may include an input surface and an input object, such as a hand or stylus, that may be utilized by a user to provide an input via the input surface. In some cases, the input may be provided without the input object touching the input surface. The location of the input object with respect to the input surface over time and a speed of the input object over time may be used to determine whether an input has been indicated by the input object without contacting the input surface.

Detection of an input object on or near a surface

Detection of human hands within an environment is used for gesture recognition for input to a computing system. Detecting a hand is non-trivial, however, because the appearances of a hand vary widely with many independent digits and moving parts. A feature-based similarity measurement is used to detect two specific hand positions&#x2014;an open hand and a closed hand.

Open and closed hand detection

Features are disclosed for an automatic segmentation and alignment of images for display via an interface. The images may have different scales and lengths. As such, items shown in the images, such as clothing, may not be depicted in a uniform way. Segmentation of the images into image portions where a portion of an image shows a specific item is described. The segmentation may be achieved using models and/or complex image analysis. To provide a realistic view of the subject when image segments are presented together on an interface, additional alignment of the image segments may be performed. The alignment may be achieved using models and/or complex image analysis.

Browsing interface for item counterparts having different scales and lengths

Techniques for detecting a hand of a user and a pose of the hand are described herein. In some instances, as a user interacts with a computing system through various hand signals, the hand of the user and pose of the hand (e.g., position and orientation) may be detected. To detect a hand, a point set may be analyzed to identify local extremum points (e.g., peaks and valleys) along an outer boundary of the point set. The local extremum points may then be used to identify a potential hand segment including one or more potential finger segments and/or a potential palm segment. When the potential hand segment satisfies one or more criteria, a hand may be detected. To detect a hand pose, an observed pose of the hand segment may be compared to one or more hand pose models.

Hand and hand pose detection

Techniques are described for performing automatic focusing of a projected image on a mobile projection surface. Specific regions of the projected image are identified that are likely to be desired to stay in optimal focus, and attributes of those specific regions, such as sharpness and contrast, can be used to determine the need to refocus the image. Advanced knowledge of the image data being projected can be utilized to determine the specific regions of the projected image that require monitoring for optimal focusing.

Automatic projection focusing

A system may utilize sound localization techniques, such as time-difference-of-arrival techniques, to estimate an audio-based sound source position from which a sound originates. An optical image or depth map of an area containing the sound source location may then captured and analyzed to detect an object that is known or expected to have produced the sound. The position of the object may also be determined based on the analysis of the optical image or depth map. The position of the sound source may then be determined based at least in part on the position of the detected object or on a combination of the audio-based sound source position and the determined position of the object.

Audio and depth based sound source localization

A hand gesture may be characterized mathematically as a set of motion parameters applied to a dynamic motion model. Training may be conducted to compile a library of motion parameter sets corresponding to various gestures. Motion parameters corresponding to observed gestures may than be compared to the library of motion parameter sets to classify the observed gestures.

Gesture recognition

A virtual user interface in a 3D environment is projected onto a display object, such as an open hand. An operating object, such as a stylus or a second hand with an index finger extended, may be used to select input features of the projected user interface. In one scenario, for example, a telephone keypad may be projected onto the open hand and the index finger may be used to select a particular number on the keypad. Images of the hands may be segmented to produce contours of the finger and open hand, and these contours are used to determine the location of the finger relative to the open hand.

Obtaining input from a virtual user interface

Techniques are described for projecting content onto a display surface and for analyzing the hand of a user that is placed over the display surface, such as when the user is making a command gesture. Based on the location and known shape of the display surface, a surface depth map is calculated corresponding to the display surface. Points of the calculated surface depth map are then subtracted from respective points an observed depth map to produce a foreground depth map corresponding to the user hand. The hand may be further characterized and tracked based on the foreground depth map.

Foreground analysis on parametric background surfaces

Techniques are described for determining the pose of a surface based on a point set representing the surface. Corner models are matched to a contour of the point set using a difference minimization technique to identify corner poses of the surface. Based on the corner poses, a virtual model of the surface is created and matched to the point set to determine the pose of the surface.

Edge-based pose detection

A number of images of an environment may be obtained over time by a number of sensors. The data obtained by the sensors may be used to generate contours of objects in the environment. In some cases, hands with pointing fingers in the environment may be identified based on the contours of the objects in the environment. In particular, contours having a palm segment and a pointing finger segment may be identified as hands with pointing fingers. In an implementation, 1-dimensional curvature and distance information may be used to identify contours having a palm segment and a pointing finger segment.

Pointing finger detection

An augmented reality environment allows interaction between virtual and real objects. Wavelength modulated light with a known spectral signature available within the augmented reality environment is used to generate spectral reflectance data for objects within the environment. This reflectance data is used to determine whether objects or portions thereof are skin.

Growing community of inventors

Cupertino, CA, United States of America

Ning Yao