Hand pose estimation for machine learning based gesture recognition
The technology disclosed performs hand pose estimation on a so-called “joint-by-joint” basis. So, when a plurality of estimates for the 28 hand joints are received from a plurality of expert networks (and from master experts in some high-confidence scenarios), the estimates are analyzed at a joint level and a final location for each joint is calculated based on the plurality of estimates for a particular joint. This is a novel solution discovered by the technology disclosed because nothing in the field of art determines hand pose estimates at such granularity and precision. Regarding granularity and precision, because hand pose estimates are computed on a joint-by-joint basis, this allows the technology disclosed to detect in real time even the minutest and most subtle hand movements, such a bend/yaw/tilt/roll of a segment of a finger or a tilt an occluded finger, as demonstrated supra in the Experimental Results section of this application.
1. A method of preparing sample hand positions for training of neural network systems, the method including:
obtaining ground truth simulated stereoscopic hand images;
extracting hand boundaries for the hand images and aligning the hand boundaries with hand centers included in the hand images;
generating translated, rotated and scaled variants of the hand boundaries and applying Gaussian jittering to at least some variants;
extracting hand regions from the at least some variants of the hand boundaries to which Gaussian jittering was applied;
computing ground truth pose vectors for the hand regions; and
storing the ground truth pose vectors in tangible machine readable memory as output labels for the ground truth simulated stereoscopic hand images.
2. The method of claim 1 , further including generating three-dimensional (3D) simulated hands in mesh models and/or capsule hand skeleton models.
3. The method of claim 1 , further including generating a simulated coordinate system to determine hand position parameters of a simulated hand in three-dimensional (3D).
4. The method of claim 1 , further including generating a simulated perspective of a simulated gesture recognition system to determine hand position parameters of a simulated hand in three-dimensional (3D).
5. The method of claim 1 , wherein ground truth pose vectors are 84 dimensional representing 28 hand joints in three-dimensional (3D) space.
6. The method of claim 1 , wherein ground truth simulated stereoscopic hand images are generated using a computer graphic simulator.
7. The method of claim 6 , wherein generating ground truth simulated stereoscopic hand images using a computer graphic simulator further includes:
accessing simulation parameters that specify at least one of:
a range of hand positions and position sequences,
a range of hand anatomies, including palm size, fattiness, stubbiness, and skin tone, and
a range of backgrounds;
accessing a camera perspective specification that specifies one or more of:
a focal length,
a field of view,
a wavelength sensitivity, and
artificial lighting conditions; and
generating a plurality of hand position-hand anatomy-background simulations, each simulation labeled with hand position parameters, the simulations organized in sequences.
8. The method of claim 7 , wherein extracting hand boundaries for the hand images further includes:
applying the camera perspective specification to render from the simulations at least a corresponding set of simulated hand position images.
9. The method of claim 8 , further including:
saving simulated hand position images as labelled with hand position parameters from corresponding simulations for use in training a hand position recognition system.
10. The method of claim 8 , wherein simulated hand position images are stereoscopic images with depth map information.
11. The method of claim 8 , wherein simulated hand position images are binocular pairs of images.
12. The method of claim 8 , wherein hand position parameters are a plurality of joint locations in three-dimensional (3D) space.
13. The method of claim 8 , wherein hand position parameters are a plurality of joint angles in three-dimensional (3D) space.
14. The method of claim 8 , wherein hand position parameters are a plurality of hand skeleton segments in three-dimensional (3D) space.
15. A non-transitory computer readable storage medium impressed with computer program instructions to prepare sample hand positions for training of neural network systems, which instructions, when executed on a processor, implement a method comprising:
obtaining ground truth simulated stereoscopic hand images;
extracting hand boundaries for the hand images and aligning the hand boundaries with hand centers included in the hand images;
generating translated, rotated and scaled variants of the hand boundaries and applying Gaussian jittering to at least some variants;
extracting hand regions from the at least some variants of the hand boundaries to which Gaussian jittering was applied;
computing ground truth pose vectors for the hand regions; and
storing the ground truth pose vectors in tangible machine readable memory as output labels for the ground truth simulated stereoscopic hand images.
16. The non-transitory computer readable storage medium of claim 15 , wherein obtaining ground truth simulated stereoscopic hand images further includes using a computer graphic simulator:
accessing simulation parameters that specify at least one of:
a range of hand positions and position sequences,
a range of hand anatomies, including palm size, fattiness, stubbiness, and skin tone, and
a range of backgrounds;
accessing a camera perspective specification that specifies one or more of:
a focal length,
a field of view,
a wavelength sensitivity, and
artificial lighting conditions; and
generating a plurality of hand position-hand anatomy-background simulations, each simulation labeled with hand position parameters, the simulations organized in sequences.
17. The non-transitory computer readable storage medium of claim 16 , wherein extracting hand boundaries for the hand images further includes:
applying the camera perspective specification to render from the simulations at least a corresponding set of simulated hand position images.
18. The non-transitory computer readable storage medium of claim 17 , wherein the simulated hand position images are selected from: stereoscopic images with depth map information and binocular pairs of images.
19. The non-transitory computer readable storage medium of claim 17 , wherein hand position parameters are selected from a plurality of joint locations in three-dimensional (3D) space, a plurality of joint angles in three-dimensional (3D) space, and a plurality of hand skeleton segments in three-dimensional (3D) space.
20. A method of preparing sample hand positions for training of neural network systems, the method including:
obtaining ground truth simulated stereoscopic hand images;
computing ground truth pose vectors for one or more hand regions extracted from the at least some translated, rotated and scaled Gaussian-jittered variants of hand-center aligned hand boundaries extracted from the ground truth simulated stereoscopic hand images; and
storing the ground truth pose vectors in tangible machine readable memory as output labels for the ground truth simulated stereoscopic hand images.