Path recognition method using a combination of invariant positional data and attributes of variation, path recognition device, path recognition program, and path recognition program recording medium
The shape or movement of a gesturing body or portion thereof in two- or three-dimensional space is ascertained from the path of the outline of the shape or the path of the movement. The method disclosed involves receiving input of data that represents a path and using artificial intelligence to recognize the meaning of the path, i.e., to recognize which of a plurality of pre-prepared meanings is the meaning of a gesture. As pre-processing for inputting location data for a point group along the path to the artificial intelligence, at least one attribute from among the location, size, and direction of the entire point group is extracted, and location data for the point group is converted to attribute invariant location data that is relative to the extracted attribute(s) but not dependent on the extracted attribute(s). Then data that includes the attribute invariant location data and the extracted attribute(s) is inputted to the artificial intelligence as input data. The result is efficient and effective processing of gesture-based dialogue between a user and a computer.
1. A path recognition method for identifying path sets including of one or more paths in 2-dimensional or 3-dimensional space by using an artificial intelligence, the method comprising:
accepting input data including path set data representing the path sets;
based on the accepted input data path set data, extracting at least one attribute of position, direction, and scale from each of one or more point groups including positional data of points along at least some part of one or more paths of the path sets;
calculating one or more point groups including positional data which is invariant to and not dependent on the at least one attribute by transforming the positional data of the point group of-points along at least part of a path of the one or more paths of the path sets with the corresponding at least one attribute;
entering data including the positional data of the point group derived from the point group along the paths constituting a path set into the artificial intelligence as input data;
receiving as output data of the artificial intelligence an estimate of the probability of the path set belonging to each of a plurality of predefined categories, the estimates being calculated by the artificial intelligence;
the input data entered into the artificial intelligence including the attribute invariant point group data of positional data of points along at least a part of one or more paths of the path set; and
the input data entered into the artificial intelligence further including at least one of the attributes of the at least some part of the one or more paths of the path sets.
2. The path recognition method of claim 1 , wherein operation mode selection data is accepted to select either a training mode or a recognition mode as mode of operation, where
when training mode is selected as an operation mode the correct output data desired from the artificial intelligence is provided as an input which is accepted as training data and by providing this training data to the artificial intelligence used to train the artificial intelligence, and
when recognition mode is selected as operation mode the output data includes, as recognition result data, a category indicated by the artificial intelligence output data to have a highest probability estimate among a plurality of categories.
3. The path recognition of claim 2 , wherein the result data set includes at least part of the attributes which have been extracted from the at least some part of the one or more paths of the path set.
4. The path recognition method according to claim 1 , wherein an average position of a whole point group of the at least part of the one or more paths of the path set is among extracted attributes as the position of the point group.
5. The path recognition method according to claim 1 , wherein a standard deviation or set of standard deviations of the whole point group of the at least part of the one or more paths of the path set is among extracted attributes as a scale of the point group.
6. The path recognition method according to claim 1 , wherein a main axis according to a principal component analysis of the whole point group of the at least part of the one or more paths of the path set is among the extracted attributes as the direction of the point group.
7. The path recognition method according to claim 1 , wherein the attributes extracted from the whole point group of the at least part of the one or more paths of the path set are the position, size, and direction.
8. The path recognition method according to claim 1 , wherein the at least part of the one or more paths of the path set is all of the one or more paths.
9. The path recognition method according to claim 1 , wherein the input data for the artificial intelligence includes at least one of the extracted attributes of each of the at least part of the one or more paths of the path set.
10. The path recognition method according to claim 1 , wherein a point group including a predetermined number of points is selected along each path of the set of paths, the point group being selected to represent each respective path by a set of consecutive line segments between the point group's points to minimize the spatial difference between the respective path and a representing point group's line segments.
11. The path recognition method according to claim 10 , wherein the selection of a point group including of a predetermined number of points for each path is done by simulated annealing.
12. The path recognition method according to claim 1 , wherein the artificial intelligence is a neural network.
13. The path recognition method according to claim 1 , wherein the accepted input data which includes the path set data also includes additional data, and the additional data is included in the input data for the artificial intelligence.
14. The path recognition method according to claim 1 , wherein each path of the path set accepted as input data represents either a contour path along an outline of a gesture or a motion path along a gesture motion.
15. The path recognition method according to claim 1 , wherein the estimates of probability calculated by the artificial intelligence are obtained by repeatedly providing data as input data and selecting an artificial intelligence according to a type of the one or more paths of the path set accepted as input data and using that data as input.
16. The path recognition method according to claim 15 , wherein the type of the one or more paths of the path set includes the number of paths constituting the path set.
17. A path recognition device comprising a computer which is capable of executing the path recognition method according to claim 1 .
18. A computer-readable path recognition program which is capable of executing the path recognition method according to claim 1 .
19. A computer-readable intransient storage medium which stores a program which is capable of executing the path recognition method according to claim 1 .