Improving predictive information for free space gesture control and communication
The technology disclosed relates to simplifying updating of a predictive model using clustering observed points. In particular, it relates to observing a set of points in 3D sensory space, determining surface normal directions from the points, clustering the points by their surface normal directions and adjacency, accessing a predictive model of a hand, refining positions of segments of the predictive model, matching the clusters of the points to the segments, and using the matched clusters to refine the positions of the matched segments. It also relates to distinguishing between alternative motions between two observed locations of a control object in a 3D sensory space by accessing first and second positions of a segment of a predictive model of a control object such that motion between the first position and the second position was at least partially occluded from observation in a 3D sensory space.
1. A computer implemented method of updating a predictive model using clustering observed points, the method including:
receiving observation information capturing a set of points in a three-dimensional (3D) sensory space;
clustering the points by their adjacency;
accessing a predictive model of a hand;
matching the clusters of the points to segments of the predictive model; and
using the matched clusters to refine positions of matched segments.
2. The method of claim 1 , further including:
selecting a reference vector and determining a difference in angle between surface normal directions determined from the points and the reference vector; and
using a magnitude of the difference to cluster the points.
3. The method of claim 2 , wherein the reference vector is orthogonal to a field of view of camera used to capture the points on an image.
4. The method of claim 2 , wherein the reference vector is along a longitudinal axis of the hand.
5. The method of claim 2 , wherein the reference vector is along a longitudinal axis of a portion of the hand.
6. The method of claim 1 , further including calculating an error indication by:
determining whether the points and points on the segments of the predictive model are within a threshold closest distance.
7. The method of claim 1 , further including calculating an error indication by:
pairing the points in the set with points on axes of the segments of the predictive model, wherein the points in the set of points lie on vectors that are normal to the axes; and
determining a reduced root mean squared deviation (RMSD) of distances between paired point sets.
8. The method of claim 1 , further including calculating an error indication by:
pairing the points in the set with points on the segments of the predictive model, wherein normal vectors to the points in the set are parallel to each other; and
determining a reduced root mean squared deviation (RMSD) of distances between bases of the normal vectors.
9. The method of claim 1 , further including:
determining physical proximity between points in the set based on the matched clusters;
based on the determined physical proximity, identifying co-located segments of the predictive model that change positions together; and
refining positions of segments of the predictive model responsive to the co-located segments.
10. The method of claim 9 , wherein the co-located segments represent adjoining figures of the hand.
11. The method of claim 9 , wherein the co-located segments represent subcomponents of a same finger.
12. A computer implemented method of distinguishing between alternative motions between two observed locations of a control object in a three-dimensional (3D) sensory space, the method including:
accessing a first position and a second position of a segment of a control object, wherein motion between the first position and the second position was at least partially occluded from observation in a three-dimensional (3D) sensory space;
receiving two or more alternative interpretations of movement from the first position to the second position;
estimating entropy or extent of motion involved in the alternative interpretations;
selecting an alternative interpretation with lower entropy or extent of motion than other interpretations; and
providing the alternative interpretation selected.
13. The method of claim 12 , wherein the control object is a hand.
14. The method of claim 12 , wherein the control object is a tool.
15. A system enabling updating a predictive model using clustering observed points, comprising:
at least one camera oriented towards a field of view;
a gesture database comprising a series of electronically stored records, each of the records relating a predictive model of a hand; and
an image analyzer coupled to the camera and the database and configured to:
observe a set of points in a three-dimensional (3D) sensory space using at least one image captured by the camera;
cluster the points by their adjacency;
access a particular predictive model of the hand;
match the clusters of the points to segments of the predictive model; and
use the matched clusters to refine positions of matched segments.
16. The system of claim 15 , further configured to:
select a reference vector and determine a difference in angle between surface normal directions determined from the points and the reference vector; and
use a magnitude of the difference to cluster the points.
17. The system of claim 16 , wherein the reference vector is orthogonal to the field of view.
18. The system of claim 16 , wherein the reference vector is along a longitudinal axis of the hand.
19. The system of claim 16 , wherein the reference vector is along a longitudinal axis of a portion of the hand.
20. A non-transitory computer readable medium storing processor executable instructions to distinguish between alternative motions between two observed locations of a control object in a three-dimensional (3D) sensory space, which instructions when executed by a processor implement actions including:
accessing a first position and a second position of a segment of a control object, wherein motion between the first position and the second position was at least partially occluded from observation in a three-dimensional (3D) sensory space;
receiving two or more alternative interpretations of movement from the first position to the second position;
estimating entropy or extent of motion involved in the alternative interpretations;
selecting an alternative interpretation with lower entropy or extent of motion than other interpretations; and
providing the alternative interpretation selected.