Creating data shapes for pattern recognition systems
Methods, apparatuses and systems directed to pattern identification and pattern recognition. In some particular implementations, the invention provides a flexible pattern recognition platform including pattern recognition engines that can be dynamically adjusted to implement specific pattern recognition configurations for individual pattern recognition applications. In some implementations, the present invention also provides for a partition configuration where knowledge elements can be grouped and pattern recognition operations can be individually configured and arranged to allow for multi-level pattern recognition schemes.
1. A computer-implemented method, comprising:
receiving input data;
generating a plurality of feature sets from the input data, each feature set representing a corresponding one of a plurality of portions of the input data, each feature set also representing a point in a corresponding multi-dimensional space;
generating a plurality of data objects from the feature sets, each data object corresponding to one or more of the feature sets, each data object representing a shape that includes the one or more points that the corresponding one or more feature sets represent; and
associating identification data with each data object, the identification data for each data object including a reference to each portion of the input data to which the data object corresponds.
2. The method of claim 1 , wherein the identification data for each data object also identifies a source of each portion of the input data to which the data object corresponds.
3. The method of claim 1 , wherein the correspondence between each data object and each of the corresponding one or more feature sets is established in a supervised learning operation, an unsupervised learning operation, or a recognition operation.
4. The method of claim 1 , wherein each data object is a knowledge element of a knowledge base, the knowledge elements and the knowledge base being configured to support pattern recognition operations, the method further comprising performing a first pattern recognition operation using additional input data by matching a feature set representing the additional input data to one or more of the knowledge elements.
5. The method of claim 4 , wherein the knowledge elements of the knowledge base are organized in a plurality of partitions, each partition being configured to support a corresponding subset of the pattern recognition operations.
6. The method of claim 5 , further comprising combining pattern recognition results from more than one of the partitions to generate a higher-level result.
7. The method of claim 1 , further comprising maintaining a count for each data object, each count representing a number of the feature sets corresponding to that data object, the counts collectively representing statistical regularities of patterns represented by the input data.
8. The method of claim 1 , wherein generating a first data object is done using one or more learning operations, and wherein the identification data associated with the first data object identifies training data used in the one or more learning operations.
9. The method of claim 1 , wherein different subsets of the feature sets include different numbers of features.
10. The method of claim 1 , wherein the input data represent signals generated by one or more sensors, and wherein the identification data associated with each data object identifies either or both of a corresponding sensor or corresponding sensor type.
11. The method of claim 1 , wherein the input data represent preexisting digital data.
12. The method of claim 1 , wherein the receiving, generating, generating, and associating are all performed by hardware or software that is included in a system that also generates the input data.
13. The method of claim 1 , wherein a first portion of the input data represents training data, and wherein generating the data objects includes:
in a learning operation, associating a first feature set corresponding to the first portion of the input data with a first data object; and
storing an association between the training data and the first data object;
the method further comprising:
in a recognition operation, matching a second feature set corresponding to a second portion of the input data to the first data object; and
in response to the second feature set matching the first data object, retrieving the training data based on the association between the training data and the first data object.
14. The method of claim 1 , wherein generating the data objects includes:
in a learning operation, associating a first feature set corresponding to a first portion of the input data with a first data object of a knowledge base;
storing an association between the first data object and remote data stored in a remote database separate from the knowledge base;
the method further comprising:
in a recognition operation, matching a second feature set corresponding to a second portion of the input data to the first data object; and
in response to the second feature set matching the first data object, retrieving the remote data from the remote database based on the association between the remote data and the first data object.
15. A system, comprising memory and one or more processors configured to:
receive input data;
generate a plurality of feature sets from the input data, each feature set representing a corresponding one of a plurality of portions of the input data, each feature set also representing a point in a corresponding multi-dimensional space;
generate a plurality of data objects from the feature sets, each data object corresponding to one or more of the feature sets, each data object representing a shape that includes the one or more points that the corresponding one or more feature sets represent; and
associate identification data with each data object, the identification data for each data object including a reference to each portion of the input data to which the data object corresponds.
16. The system of claim 15 , wherein the metadata identification data for each data object also identifies a source of each portion of the input data to which the data object corresponds.
17. The system of claim 15 , wherein the correspondence between each data object and each of the corresponding one or more feature sets is established in a supervised learning operation, an unsupervised learning operation, or a recognition operation.
18. The system of claim 15 , wherein each data object is a knowledge element of a knowledge base, the knowledge elements and the knowledge base being configured to support pattern recognition operations, and wherein the memory and one or more processors are further configured to perform a first pattern recognition operation using additional input data by matching a feature set representing the additional input data to one or more of the knowledge elements.
19. The system of claim 18 , wherein the knowledge elements of the knowledge base are organized in a plurality of partitions, each partition being configured to support a corresponding subset of the pattern recognition operations.
20. The system of claim 19 , wherein the memory and one or more processors are further configured to combine pattern recognition results from more than one of the partitions to generate a higher-level result.
21. The system of claim 15 , wherein the memory and one or more processors are further configured to maintain a count for each data object, each count representing a number of the feature sets corresponding to that data object, the counts collectively representing statistical regularities of patterns represented by the input data.
22. The system of claim 15 , wherein the memory and one or more processors are configured to generate a first data object using one or more learning operations, and wherein the identification data associated with the first data object identifies training data used in the one or more learning operations.
23. The system of claim 15 , wherein different subsets of the feature sets include different numbers of features.
24. The system of claim 15 , wherein the input data represent signals generated by one or more sensors, and wherein the identification data associated with each data object identifies either or both of a corresponding sensor or corresponding sensor type.
25. The system of claim 15 , wherein the input data represent preexisting digital data.
26. The system of claim 15 , wherein the system is also configured to generate the input data.
27. The system of claim 15 , wherein a first portion of the input data represents training data, and wherein the memory and one or more processors are configured to generate the data objects by:
in a learning operation, associating a first feature set corresponding to the first portion of the input data with a first data object; and
storing an association between the training data and the first data object; wherein the memory and one or more processors are further configured to
in a recognition operation, matching a second feature set corresponding to a second portion of the input data to the first data object; and
in response to the second feature set matching the first data object, retrieving the training data based on the association between the training data and the first data object.
28. The system of claim 15 , wherein the memory and one or more processors are configured to generate the data objects by:
in a learning operation, associating a first feature set corresponding to a first portion of the input data with a first data object of a knowledge base;
storing an association between the first data object and remote data stored in a remote database separate from the knowledge base;
wherein the memory and one or more processors are further configured to:
in a recognition operation, matching a second feature set corresponding to a second portion of the input data to the first data object; and
in response to the second feature set matching the first data object, retrieving the remote data from the remote database based on the association between the remote data and the first data object.