Depth-value classification using forests
Human Computer Interfaces (HCI) may allow a user to interact with a computer via a variety of mechanisms, such as hand, head, and body gestures. Various of the disclosed embodiments allow information captured from a depth camera on an HCI system to be used to recognize such gestures. Particularly, by training a classifier using vectors having both base and extended components, more accurate classification results may be subsequently obtained. The base vector may include a leaf-based assessment of the classification results from a forest for a given depth value candidate pixel. The extended vector may include additional information, such as the leaf-based assessment of the classification results for one or more pixels related to the candidate pixel. Various embodiments employ this improved structure with various optimization methods and structure to provide more efficient in-situ operation.
1. A computer system configured to generate a depth-value classifier comprising:
at least one processor;
at least one memory comprising instructions configured to cause the at least one processor to perform a method comprising:
receiving a candidate depth pixel associated with a class;
generating a base vector by applying the candidate depth pixel to one or more trees;
generating an extended vector;
providing the base vector and the extended vector to a machine learning system to generate a classifier;
wherein generating an extended vector comprises:
determining classification probabilities for nodes in trees of the one or more trees for a related pixel; and
generating a vector based on the classification probabilities for the nodes, wherein the number of entries in the extended vector is less than the number of entries in the base vector.
2. The computer system of claim 1 , wherein generating an extended vector comprises concatenating base vectors of one or more pixels neighboring the candidate depth pixel.
3. The computer system of claim 1 , wherein generating an extended vector comprises summing base vectors of one or more pixels neighboring the candidate depth pixel.
4. The computer system of claim 3 , wherein
the machine learning system is a support vector machine, wherein
generating an extended vector comprises performing a second pass after generating the base vector, and wherein
providing the base vector and the extended vector comprises providing a combined vector, the combined vector comprising the extended vector appended to the base vector.
5. The computer system of claim 4 , the method further comprising:
determining that a classification with the classifier was successful;
generating new training data in association with the successful classification; and
generating a new classifier using the new training data.
6. The computer system of claim 1 ,
wherein the machine learning system is a support vector machine and wherein the method further comprises determining that the number of entries in the extended vector will include less than the number of entries in the base vector based, at least in part, on a determination that the system is still engaged in a classification session.
7. The computer system of claim 1 , wherein
the machine learning system is a support vector machine, and wherein
providing the base vector and the extended vector comprises providing a combined vector, the combined vector comprising the extended vector appended to the base vector.
8. A computer-implemented method comprising:
receiving a candidate depth pixel associated with a class;
generating a base vector by applying the candidate depth pixel to one or more trees;
generating an extended vector;
providing the base vector and the extended vector to a machine learning system to generate a classifier;
wherein generating an extended vector comprises:
determining classification probabilities for nodes in trees of the one or more trees for a related pixel; and
generating a vector based on the classification probabilities for the nodes, wherein the number of entries in the extended vector is less than the number of entries in the base vector.
9. The computer-implemented method of claim 8 , wherein generating an extended vector comprises concatenating base vectors of one or more pixels neighboring the candidate depth pixel.
10. The computer-implemented method of claim 8 , wherein generating an extended vector comprises summing base vectors of one or more pixels neighboring the candidate depth pixel.
11. The computer-implemented method of claim 10 , wherein
the machine learning system is a support vector machine, wherein
generating an extended vector comprises performing a second pass after generating the base vector, and wherein
providing the base vector and the extended vector comprises providing a combined vector, the combined vector comprising the extended vector appended to the base vector.
12. The computer-implemented method of claim 11 , the method further comprising:
determining that a classification with the classifier was successful;
generating new training data in association with the successful classification; and
generating a new classifier using the new training data.
13. The computer-implemented method of claim 8 , wherein the machine learning system is a support vector machine and wherein the method further comprises determining that the number of entries in the extended vector will include less than the number of entries in the base vector based, at least in part, on a determination that the system is still engaged in a classification session.
14. The computer-implemented method of claim 8 , wherein
the machine learning system is a support vector machine, and wherein
providing the base vector and the extended vector comprises providing a combined vector, the combined vector comprising the extended vector appended to the base vector.
15. A non-transitory computer-readable medium comprising instructions configured to cause a computer system to perform a method comprising:
receiving a candidate depth pixel associated with a class;
generating a base vector by applying the candidate depth pixel to one or more trees;
generating an extended vector;
providing the base vector and the extended vector to a machine learning system to generate a classifier;
wherein generating an extended vector comprises:
determining classification probabilities for nodes in trees of the one or more trees for a related pixel; and
generating a vector based on the classification probabilities for the nodes, wherein the number of entries in the extended vector is less than the number of entries in the base vector.
16. The non-transitory computer-readable medium of claim 15 , wherein generating an extended vector comprises concatenating base vectors of one or more pixels neighboring the candidate depth pixel.
17. The non-transitory computer-readable medium of claim 15 , wherein generating an extended vector comprises summing base vectors of one or more pixels neighboring the candidate depth pixel.
18. The non-transitory computer-readable medium of claim 17 , wherein
the machine learning system is a support vector machine, wherein
generating an extended vector comprises performing a second pass after generating the base vector, and wherein
providing the base vector and the extended vector comprises providing a combined vector, the combined vector comprising the extended vector appended to the base vector.
19. The non-transitory computer-readable medium of claim 18 , the method further comprising:
determining that a classification with the classifier was successful;
generating new training data in association with the successful classification; and
generating a new classifier using the new training data.
20. The non-transitory computer-readable medium of claim 15 , wherein the machine learning system is a support vector machine and wherein the method further comprises determining that the number of entries in the extended vector will include less than the number of entries in the base vector based, at least in part, on a determination that the system is still engaged in a classification session.
21. The non-transitory computer-readable medium of claim 15 , wherein
the machine learning system is a support vector machine, and wherein
providing the base vector and the extended vector comprises providing a combined vector, the combined vector comprising the extended vector appended to the base vector.