Abstract: Methods and systems for touchless control of a device include receiving an image from a camera and applying a machine learning technique the image to determine a first map, the first map including a function that generates a correspondence between data associated with the image to a feature space. A second map including a representation of a user in the image can be determined based on the first map. Further, a pose of the user can be determined based on the second map. The pose of the user and a portion of a body of the user can be tracked to determine a mapping between the portion of the user's body and an input feature of a device. The input feature can be modified based on the mapping to enable a touchless interaction between the user and the device.

Abstract: There is provided a computer-implemented method of training a speech-to-speech (S2S) machine learning (ML) model for adapting at least one voice attribute of speech, comprising: creating an S2S training dataset of a plurality of S2S records, wherein an S2S record comprises: a first audio content comprising speech having at least one first voice attribute, and a ground truth label of a second audio content comprising speech having at least one second voice attribute, wherein the first audio content and the second audio content have the same lexical content and are time-synchronized, and training the S2S ML model using the S2S training dataset, wherein the S2S ML model is fed an input of a source audio content with at least one source voice attribute and generates an outcome of the source audio content with at least one target voice attribute.

Abstract: A system and method for implementing a virtual fitness application are disclosed. Embodiments may be built for various platforms, including web browsers and mobile phones. One embodiment uses pose estimation to count the repetitive motions of a fitness activity (e.g., running). In one embodiment, to count the repetitive motions, a repetitive motion counting process is used that is based on computing differences of Y-coordinates of key points of the user. The repetitive motion counting process includes selecting a proper pose of the user; computing one or more delta values (corresponding to Y-coordinate changes of the key points); and counting a given user movement as a repetitive motion based on a function (e.g., average) of the delta values. One embodiment uses the pose estimation and repetitive motion count results to provide a gamified experience, for example, a leaderboard, a stats report, instant visual feedback, badges, coins, and a social experience.

Abstract: Methods and systems for image clustering are described, and include first receiving a plurality of images from a video of a game. Next, determining a first feature vector of a first player from the plurality of images by inputting the plurality of images to an artificial intelligence (AI)-based process. Next, dividing the plurality of images into a first subset of images and a second subset of images based on the first feature vector. Next, receiving a first identifier tagging the first player in a first image of the first subset. Finally, identifying a second player in a second image in the second subset by propagating the first identifier of the first subset, based on a distance measure associated with the first feature vector from the artificial intelligence (AI)-based process. Running on a mobile computing device, the method allows multiplayer tagging to be easily performed in almost any environment.

Abstract: There is provided a computer-implemented method of training a speech-to-speech (S2S) machine learning (ML) model for adapting at least one voice attribute of speech, comprising: creating an S2S training dataset of a plurality of S2S records, wherein an S2S record comprises: a first audio content comprising speech having at least one first voice attribute, and a ground truth label of a second audio content comprising speech having at least one second voice attribute, wherein the first audio content and the second audio content have the same lexical content and are time-synchronized, and training the S2S ML model using the S2S training dataset, wherein the S2S ML model is fed an input of a source audio content with at least one source voice attribute and generates an outcome of the source audio content with at least one target voice attribute.

Abstract: A system for scoring and standard analysis of user responses to an assessment test, wherein the system includes a scoring engine having one or more rubric items used to score and assess a candidate’s response to one or more free-text questions. A candidate’s response can be input into the scoring engine and optionally in communication with a machine learning classifier can produce one or more outputs. The outputs can include a score, recommendation, and user feedback among other things. The system can further include one or more machine learning classifier engines.

Abstract: Methods and systems for multi-player training of body-eye coordination using mobile computing devices each having a camera are provided. The methods and systems execute instructions to capture a first training video of a first of player, using a first camera on a first mobile computing device; superimpose a visual cue onto the first training video at a visual cue location and for a cue period starting from a first time instant; determine whether the first player has responded to the visual cue at a second time instant within the cue period; receive a notification from a second mobile computing device, where the notification was generated in response to determining that a second player has responded to the visual cue at a third time instant within the cue period; and in response to a determination that the first player has responded to the visual cue and to the notification, generate a feedback to the first player.

Abstract: Methods and systems for real-time generation of ball shot analytics are disclosed. The methods and systems perform steps for ball and posture detection, ball and posture flow generation, shot event identification and classification, and shot analytics generation based on identified shot events and shooter posture flow. Embodiments of the present invention use computer vision techniques to enable a resource-limited mobile device such as a smartphone to conduct the aforementioned steps. Also disclosed are benefits of the new methods, and alternative embodiments of implementation.

Abstract: Methods and systems for determining a ball shot attempt location on a ball court are disclosed. The method includes receiving an input video, where the input video includes the ball shot attempt and one or more players; receiving a declared ball shot attempt; determining a potential shooter by detecting a moving object flow in the input video; determining a potential shooter image region in the image plane by analyzing an overlapping image area of the ball trajectory and the potential shooter; determining a shooter foot-on-court time; determining a most probable pair of shooter legs in the potential shooter image region; determining a shooter foot location based on the most probable pair of shooter legs; and determining a ball shot attempt location on the ball court by applying a camera projection to the shooter foot location.

Abstract: Methods and systems for constructing a three-dimensional (3D) model of a user in a virtual environment for full body virtual reality (VR) applications are described. The method includes receiving an image of the user captured using an RGB camera; detecting a body bounding box associated with the user using a first trained neural network; determining a segmentation map of the user, based on the body bounding box; determining a two-dimensional (2D) contour of the user from the segmentation map; forming a 3D extrusion model by extruding the 2D contour; and constructing the 3D model of the user in the virtual environment by applying a geometric transformation to the 3D extrusion model. Applications of full body VR include physical training and fitness sessions, games, control of computing devices, manipulation and display of data, interactive social media with VR, and the like.

Abstract: Methods and systems for determining and classifying a number of repetitive motions in a video are described, and include the steps of first determining a plurality of images from a video, where the images are segmented from at least one video frame of the video. Next, performing a pose detection process on a feature of the images to generate one or more landmarks. Next, determining one or more principle component axes on points associated with a given landmark. Finally, determining at least one repetitive motion based on a pattern associated with a projection of the points onto the one or more principle components. In some embodiments, the disclosed methods can classify the repetitive motions to respective types. The present invention can be implemented for convenient use on a mobile computing device, such as a smartphone, for tracking exercises and similar repetitive motions.

Abstract: Methods and systems for determining a ball shot attempt result, the ball shot attempt captured by a camera on a mobile computing device, are disclosed. The methods and systems perform steps comprising receiving an input video captured using the camera on the mobile computing device, wherein the input video comprises a ball shot attempt at a goal; receiving a declared ball shot attempt, the declared ball shot attempt detected by analyzing a ball trajectory in the input video; detecting an element movement associated with an element of the goal; and determining the ball shot attempt result based on the element movement and the ball trajectory. In some embodiments, the element of the goal is a net, and the declared ball shot attempt is detected by analyzing the ball trajectory in a ball shot attempt Region of Interest (ROI) within frames of the input video.

Abstract: Methods and systems for remote multiplayer gameplay are disclosed. An example method can include joining a multiplayer interactive physical game including first and second players competing in an activity; capturing a first game video of a first player in a first play of the multiplayer interactive physical game, using a first camera on a first mobile computing device in a first gaming area; determining a first object flow associated with the first player from the first game video, by performing an algorithm on one or more frames of the first game video to detect one or more key points of the first player; generating a first analytic based on the first object flow; receiving a second analytic associated with the second player in a second gaming area; and generating a feedback to the first player based on the first analytic or the second analytic.

Abstract: A first method and system for customizing a file comprising at least one of audio or video content comprises receiving a baseline file comprising at least one of audio or video content, receiving a first file, and generating a customized file comprising at least one of audio or video content comprising the first file and the baseline file comprising at least one of audio or video content. A second method and system for customizing a media file comprises receiving a first baseline media file, receiving a first customization file, generating a first customized media file comprising the baseline media file and the first customization file, determining one or more customization preferences, and customizing a second baseline media file using the one or more customization preferences to generate a customized second media file in response to a triggering event.

Abstract: Methods and systems for image clustering are described, and include determining a plurality of images from a video of a game, the video captured by a camera on a mobile device, where at least one image of the plurality of images is segmented from a video frame of the video; determining a feature vector from the at least one image; dividing the images into a first subset and a second subset based on the feature vector; tagging a first player in a first image of the first subset with an identifier, where the identifier differentiates the images in the first subset to a plurality of players; and identifying a second player in a second image in the second subset by propagating the identifier of the first subset, based on a distance measure associated with the feature vector. Running on a mobile computing device, this invention allows multiplayer tagging to be easily performed in almost any environment.

Abstract: Methods and systems for 3D ball trajectory reconstruction are disclosed. The methods and systems perform steps of receiving a 2D ball trajectory, a shooter's on-ground foot location, and a camera projection matrix; generating on-ground trajectory line assumptions based on the shooter's on-ground foot location and the 2D trajectory; reconstructing a candidate 3D trajectory for each on-ground trajectory line assumption, by determining a ball height for each on-ground projection point in the on-ground trajectory line and by curve fitting; evaluating each candidate 3D trajectory and selecting one with a best score as the reconstructed 3D ball trajectory. Embodiments of the present invention enable a resource-limited mobile device such as a smartphone to conduct the aforementioned steps. Also disclosed are benefits of the new methods, and alternative embodiments of implementation.

Abstract: Methods and systems for performing location and movement determination in a three-dimensional environment using a mobile device with a camera are disclosed. The methods and systems perform the steps of capturing one or more images of a user using the camera on the mobile computing device, determining the user from the images using a machine learning algorithm, determining a pose associated with the user using a machine vision algorithm, determining a depth of the user with respect to the camera based on the pose and an input height of the user using a binary search process, and determining the location of the user in the environment based on the pose and the depth.

Abstract: Methods and systems for virtual coaching and performance training using a mobile device are disclosed. The methods and systems perform the steps of capturing a training video of one or more players, using the camera on the mobile computing device; superimposing a visual cue onto the training video at a first location and for a cue period starting from a first time instant; determining whether at least one of the one or more players has responded to the visual cue at a second time instant within the cue period, by analyzing a body posture flow of each player between the first time instant and the second time instant, wherein each body posture flow is extracted from the training video by performing a computer vision algorithm on one or more frames of the training video; and in response to determining that at least one player has responded to the visual cue, generating a feedback to the one or more players.

Abstract: A gearless differential for distributing power from a motor to two follower shafts with ends connected to opposite wheels and to opposed outer disks held in position with a journal mounted casing having corresponding identical regular polygonal raceways with asymmetric sections surrounding a central drive disk there between with equidistant radial slots each holding one cylinder or roller bearing to radially reciprocate as they race along the raceways in a slippage mode where one wheel loses traction; and in a drive mode where the central drive disk power input drives both outer disks connected to the two follower shafts transferring torque evenly to both wheels so there is no movement of the cylinders or roller bearings in the raceways.