CPC G06V 40/23 (2022.01) [A63B 24/0006 (2013.01); A63B 24/0062 (2013.01); A63B 24/0087 (2013.01); A63B 69/3605 (2020.08); A63B 71/0622 (2013.01); G06T 13/40 (2013.01); G06V 10/34 (2022.01); A63B 2024/0009 (2013.01); A63B 2024/0015 (2013.01); A63B 2024/0096 (2013.01); A63B 69/36 (2013.01); A63B 2071/063 (2013.01); A63B 2071/0636 (2013.01); A63B 2071/0647 (2013.01); A63B 2220/05 (2013.01); A63B 2220/806 (2013.01)] | 44 Claims |
1. A system configured for applying biomechanical analysis to a sequence of two-dimensional (2D) images of a user's movement during performance of a sport activity to
quantitatively analyze the user's movement of predetermined body parts during certain phases of the sport activity, generate a computer-generated 3D avatar of the user's performance, including movement of the predetermined body parts, the system comprising:
a data storage device for storing user-definable watchlists comprising a list of measurable attributes of the performance to be monitored during performance of the sport activity and motion trackers comprising a user-definable sequence of measurements to be taken during the performance, including specific kinematic parameters and/or sequences of motion;
one or more hardware processors configured by machine-readable instructions to:
receive, from an image capture device, the sequence of 2D images of the user's movement during the performance of the sport activity;
implement a movement module configured to quantitatively analyze the sequence of 2D images to measure and quantify the user's movement in 2D space based on the measurable attributes in a selected watchlist, including the movement of predetermined body parts specified in the selected watchlist and measurement of the kinematic parameters and/or sequences of motions corresponding to one or more selected motion trackers;
provide the sequence of 2D images, including individual frames of the sequence, to a machine learning (ML) module, where individual frames contain 2D coordinates for individual body parts of one or more limbs that are moving during the performance of the sport activity;
provide the frames to a 3D ML model;
generate, based on the 3D ML model, a set of frames with 3D coordinates;
implement an inference process on the set of frames with the 3D coordinates for generating the computer-generated 3D avatar of the user's performance based on the measured, quantified movement;
generate the computer-generated 3D avatar of the user's performance based on the measured, quantified movement and animate the computer-generated 3D avatar to cause movement of the computer-generated 3D avatar in three-dimensional space based on the measured, quantified movement from the 2D images, wherein the animation of the computer-generated 3D avatar comprises 3D pose estimation that corresponds with a model-based generative method that treats the predetermined body parts identified in the selected watchlist as an articulated structure with a model that includes the predetermined body parts and the spatial relationship between adjacent parts; and
display the computer-generated 3D avatar on a user interface with display options for enabling the user to view the computer-generated 3D avatar from any angle during certain performance of the sport activity.
|