Information classifying device, information classifying method, information classifying program, information classifying system

• US 7,693,683 B2
• Filed: 11/17/2005
• Issued: 04/06/2010
• Est. Priority Date: 11/25/2004
• Status: Expired due to Fees

• Current Assignee: Sharp Kabushiki Kaisha (Hon Hai Precision Industry Co., Ltd.)
• Original Assignee: Sharp Kabushiki Kaisha (Hon Hai Precision Industry Co., Ltd.)
• Inventors: Ihara, Masayoshi
• Primary Examiner(s): Nghiem; Michael P.
• Assistant Examiner(s): Le; Toan M
• Application Number:
  US11/791,705
• Publication Number:
  US 20080114564A1
• Time in Patent Office: 1,601 Days
• Field of Search: 702/179, 702/158, 702/181, 702/155, 702/182, 703/2, 704/231, 704/234, 704238-240, 704/243, 704/245, 382/106, 382/181, 382/224, 382/225, 707 1- 10, 707/101, 707/E17.089
• US Class Current: 702/179
• CPC Class Codes: G06F 16/2228 Indexing structures
• CPC Class Codes: G06F 16/35 Clustering; Classification
• CPC Class Codes: G06F 16/353 into predefined classes
• CPC Class Codes: G06F 18/22 Matching criteria, e.g. proximity measures
• CPC Class Codes: G06F 18/23211 with adaptive number of clusters
• CPC Class Codes: G06F 18/24137 Distances to cluster centroïds
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06V 10/761 Proximity, similarity or dissimilarity measures
• CPC Class Codes: G06V 10/763 Non-hierarchical techniques, e.g. based on statistics of modelling distributions
• CPC Class Codes: G06V 10/764 using classification, e.g. of video objects
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence
• CPC Class Codes: G16H 50/70 for mining of medical data, e.g. analysing previous cases of other patients
• CPC Class Codes: G16H 50/80 for detecting, monitoring or modelling epidemics or pandemics, e.g. flu

Abstract

An information classifying device calculates, for a plurality of populations containing pieces of sample information, evaluation distance between a center of gravity of the pieces of sample information belonging to each population and a piece of sample information as an object of classification (object sample), calculates statistical information such as mean, variance and standard deviation of the evaluation distance for each population, evaluates the evaluation distance of the sample information to the population based on the evaluation distance and the statistical information and evaluates degree of assignment relevancy of the object sample to the population, determines to which population the object sample is to be assigned in accordance with the degree of assignment relevancy, and assigns the object sample to the population. Evaluation distance between the center of gravity of each updated population and the object sample belonging to each population is calculated. If the degree of assignment relevancy to every population is out of a prescribed range, a new population is formed, and the object sample is assigned to the new population. Thus, autonomous and stable classification of object sample to a population becomes possible.

Claims

1. An information classifying device, comprising:

distance calculating means for calculating distance information between each of an arbitrary number of pieces of sample information and center of gravity of each of an arbitrary number of populations including said pieces of sample information, using a distance evaluation function;

statistical information calculating means for calculating statistical information using mean of sample distances and variance of sample distances with the mean of sample distances being the center, based on each group of distance information constituted by said distance information calculated by said distance calculating means, associated with each said sample information belonging to each said population;

degree of assignment relevancy evaluating means for evaluating degree of assignment relevancy, by calculating distance between the center of gravity of each said population and sample information as an object of classification using said distance calculating means, normalizing said distance using said statistical information calculated by said statistical information calculating means on said calculated distance and performing a statistical test;

assignment destination determining means for determining to which population said piece of information as the object of classification is to be assigned, according to the degree of assignment relevancy evaluated by said degree of assignment relevancy evaluating means; and

sample information classifying means for assigning and classifying said piece of sample information as the object of classification to the population determined by said assignment destination determining means.

2. The information classifying device according to claim 1,

wherein

said statistical information is statistical information formed by further adding standard deviation of sample distance based on each group of said distance information belonging to each said population;

and

said assignment relevancy evaluating means evaluates degree of assignment relevancy of said sample information as the object of classification to each said population, dependent on a probability of assignment to each population, obtained by calculating distance between the center of gravity of each said population and said sample information as the object of classification using said distance evaluation function by said distance calculating means, and normalizing the distance to each said sample information as the object of classification with said standard deviation using said calculated distance and said statistical information calculated by said statistical information calculating means.

3. The information classifying device according to claim 1, wherein

said distance calculating means further includes distance evaluation function reconfiguring means for reconfiguring, after said sample information classifying means forms updated populations having sample information groups updated based on said sample information as the object of classification, the distance evaluation function to be used by said distance calculating means, in accordance with classified group of sample information for each updated population thus formed;

and

said distance calculating means further calculates a group of distance information between the sample information as the object of classification belonging to each said updated population and the center of gravity of the updated population, using the distance evaluation function reconfigured by said distance evaluation function reconfiguring means.

4. An information recognizing device, comprising:

the information classifying device according to claim 3; and

recognizing means for performing a process of recognizing identification information corresponding to a feature extracted from natural information using said distance evaluation function reconfigured by said distance evaluation function reconfiguring means.

5. An information searching device, comprising:

the information classifying device according to claim 3;

recognizing means for performing a process of recognizing identification information corresponding to a feature extracted from natural information using said distance evaluation function reconfigured by said distance evaluation function reconfiguring means; and

searching means for performing a search, using result of recognition by said recognizing means.

6. The information classifying device according to claim 1, wherein

said assignment destination determining means includes population generating means for generating a new population when degree of assignment relevancy to every population is out of a prescribed range, and determines that said piece of sample information as the object of classification is assigned to the generated.

7. The information classifying device according to claim 6, wherein

said degree of assignment relevancy is deviation from the mean of distance information group for the population;

and

said prescribed degree is in a range in which said deviation is from said mean to a prescribed multiple of standard deviation.

8. The information classifying device according to claim 6, further comprising:

population removing means for removing a population that does not contain at least a prescribed number of pieces of sample information, and for assigning pieces of sample information that belonged to the removed population to another population.

9. The information classifying device according to claim 1, wherein

said assignment destination determining means determines that said piece of sample information as the object of classification is assigned to a population to which degree of assignment relevancy evaluated by said degree of assignment relevancy evaluating means is the highest.

10. The information classifying device according to claim 1, wherein

said distance calculating means calculates said distance information based on covariance structure analysis.

11. The information classifying device according to claim 1, wherein

said distance calculating means calculates said distance information based on an eigenvalue and an eigenvector.

12. The information classifying device according to claim 1, wherein

said distance calculating means calculates Mahalanobis distance as said distance information.

13. The information classifying device according to claim 1, wherein

said distance calculating means calculates distance in accordance with Bayesian discrimination function as said distance information.

14. The information classifying device according to claim 1, wherein

said distance calculating means includes distance normalizing means for normalizing the calculated distance information.

15. The information classifying device according to claim 1, wherein

said piece of sample information as the object of classification is an arbitrary vector information, matrix information or tensor information with an identifier allotted beforehand to each element;

said distance evaluation function is a function having the vector information, matrix information or tensor information of a prescribed element configuration with an identifier allotted beforehand to each element; and

said distance calculating means calculates said distance information by reconfiguring each element of said arbitrary vector information, matrix information or tensor information such that the identifier of each element of said arbitrary vector information, matrix information or tensor information becomes the same as the identifier of the element of said prescribed element configuration, and inputting to said distance evaluation function.

16. An information classification method executed by a computer, comprising the steps of:

calculating, by a distance information calculating device configured within the computer, distance information between each of an arbitrary number of pieces of sample information and center of gravity of each of an arbitrary number of populations including said pieces of sample information, using a distance evaluation function;

calculating, by a statistical information calculating device configured within the computer, statistical information using mean of sample distances and variance of sample distances with the mean of sample distances being the center, based on each group of distance information constituted by said calculated distance information, associated with each said sample information belonging to each said population;

evaluating, by a degree of assignment relevancy evaluating unit configured within the computer, degree of assignment relevancy by calculating distance between the center of gravity of each said population and sample information as an object of classification and normalizing said distance using said calculated statistical information on said calculated distance and performing a statistical test;

determining, by an assignment destination determining unit configured within the computer, to which population said piece of information as the object of classification is to be assigned, in accordance with the evaluated degree of assignment relevancy; and

assigning and classifying, by a sample information classifying unit configured within the computer, said piece of sample information as the object of classification to the determined population.

17. The information classifying method according to claim 16, wherein

said piece of sample information as the object of classification is an arbitrary vector information, matrix information or tensor information with an identifier allotted beforehand to each element;

said distance evaluation function is a function having the vector information, matrix information or tensor information of a prescribed element configuration with an identifier allotted beforehand to each element; and

in said step of calculating said distance information, said distance information is calculated by reconfiguring each element of said arbitrary vector information, matrix information or tensor information such that the identifier of each element of said arbitrary vector information, matrix information or tensor information becomes the same as the identifier of the element of said prescribed element configuration, and inputting to said distance evaluation function.

18. A computer-readable medium, storing a set of instructions, executed by a processor, to perform an information classifying method, comprising:

calculating distance information between each of an arbitrary number of pieces of sample information and center of gravity of each of an arbitrary number of populations including said pieces of sample information, using a distance evaluation function;

calculating statistical information using mean of sample distances and variance of sample distances with the mean of sample distances being the center, based on each group of distance information constituted by said calculated distance information, associated with each said sample information belonging to each said population;

evaluating degree of assignment relevancy by calculating distance between the center of gravity of each said population and sample information as an object of classification and normalizing said distance using said calculated statistical information on said calculated distance and performing a statistical test;

determining to which population said piece of information as the object of classification is to be assigned, in accordance with the evaluated degree of assignment relevancy; and

assigning and classifying said piece of sample information as the object of classification to the determined population.

19. The computer-readable medium according to claim 18, wherein

said piece of sample information as the object of classification is an arbitrary vector information, matrix information or tensor information with an identifier allotted beforehand to each element;

said distance evaluation function is a function having the vector information, matrix information or tensor information of a prescribed element configuration with an identifier allotted beforehand to each element; and

in said step of calculating said distance information, said distance information is calculated by reconfiguring each element of said arbitrary vector information, matrix information or tensor information such that the identifier of each element of said arbitrary vector information, matrix information or tensor information becomes the same as the identifier of the element of said prescribed element configuration, and inputting to said distance evaluation function.

20. An information classifying system, comprising:

an information classifying device, and an information terminal connected to said information classifying device through a communication line;

wherein

said information classifying device includes

population receiving means for receiving an arbitrary number of populations including pieces of sample information, from said information terminal,

distance calculating means for calculating distance information between each of an arbitrary number of pieces of sample information included in said populations received by said population receiving means and center of gravity of each of the arbitrary number of populations, using a distance evaluation function,

statistical information calculating means for calculating statistical information using mean of sample distances and variance of sample distances with the mean of sample distances being the center, based on each group of distance information constituted by said distance information calculated by said distance calculating means, associated with each said sample information belonging to each said population,

degree of assignment relevancy evaluating means for evaluating degree of assignment relevancy, by calculating distance between the center of gravity of each said population and sample information as an object of classification using said distance calculating means, and normalizing said distance using said statistical information calculated by said statistical information calculating means on said calculated distance and performing a statistical test,

assignment destination determining means for determining to which population said piece of information as the object of classification is to be assigned, according to the degree of assignment relevancy determined by said degree of assignment relevancy determining means,

sample information classifying means for assigning and classifying said piece of sample information as the object of classification to the population determined by said assignment destination determining means, and

classified population passing means for passing a classified population to which said piece of sample information as the object of classification has been assigned and classified by said sample information classifying means, to said information terminal; and

said information terminal includes

population passing means for passing said arbitrary number of populations to said information classifying device, and

classified population receiving means for receiving said classified population from said information classifying device.

22. The information classifying system according to claim 20 or 21, wherein

said piece of sample information as the object of classification is an arbitrary vector information, matrix information or tensor information with an identifier allotted beforehand to each element;

said distance evaluation function is a function having the vector information, matrix information or tensor information of a prescribed element configuration with an identifier allotted beforehand to each element; and

said distance calculating means calculates said distance information by reconfiguring each element of said arbitrary vector information, matrix information or tensor information such that the identifier of each element of said arbitrary vector information, matrix information or tensor information becomes the same as the identifier of the element of said prescribed element configuration, and inputting to said distance evaluation function.

21. An information classifying system, comprising:

an information classifying device, and an information terminal connected to said information classifying device through a communication line;

wherein

said information classifying device includes

sample information receiving means for receiving a piece of sample information as an object of classification from said information terminal,

distance calculating means for calculating distance information between each of an arbitrary number of pieces of sample information including sample information as an object of classification received by said sample information receiving means and center of gravity of each of the arbitrary number of populations including said sample information, using a distance evaluation function,

statistical information calculating means for calculating statistical information using mean of sample distances and variance of sample distances with the mean of sample distances being the center, based on each group of distance information constituted by said distance information calculated by said distance calculating means, associated with each said sample information belonging to each said population,

degree of assignment relevancy evaluating means for evaluating degree of assignment relevancy, by calculating distance between the center of gravity of each said population and sample information as an object of classification using said distance calculating means, and normalizing said distance using said statistical information calculated by said statistical information calculating means on said calculated distance and performing a statistical test,

assignment destination determining means for determining to which population said piece of information as the object of classification is to be assigned, according to the degree of assignment relevancy determined by said degree of assignment relevancy determining means, and

population identifying information passing means for passing population identifying information identifying the population determined by said assignment destination determining means to said information terminal; and

said information terminal includes

sample information passing means for passing said piece of sample information as the object of classification to said information classifying device, and

population identifying information receiving means for receiving said population identifying information from said information classifying device.

23. An information providing service, using an information classifying system including

an information classifying device, and an information terminal connected to said information classifying device through a communication line;

wherein

said information classifying device includes

sample information receiving means for receiving a piece of sample information as an object of classification from said information terminal,

distance calculating means for calculating distance information between each of an arbitrary number of pieces of sample information including sample information as an object of classification received by said sample information receiving means and center of gravity of each of the arbitrary number of populations including said sample information, using a distance evaluation function,

statistical information calculating means for calculating statistical information, using mean of sample distances and variance of sample distances with the mean of sample distances being the center, based on each group of distance information constituted by said distance information calculated by said distance calculating means, associated with each said sample information belonging to each said population,

degree of assignment relevancy evaluating means for evaluating degree of assignment relevancy, by calculating distance between the center of gravity of each said population and sample information as an object of classification using said distance calculating means, and normalizing said distance using said statistical information calculated by said statistical information calculating means on said calculated distance and performing a statistical test,

assignment destination determining means for determining to which population said piece of information as the object of classification is to be assigned, according to the degree of assignment relevancy determined by said degree of assignment relevancy determining means, and

population identifying information passing means for passing population identifying information identifying the population determined by said assignment destination determining means to said information terminal; and

said information terminal includes

sample information passing means for passing said piece of sample information as the object of classification to said information classifying device, and

population identifying information receiving means for receiving said population identifying information from said information classifying device.

24. The information providing service using the information classifying system according to claim 23, wherein

said piece of sample information as the object of classification is an arbitrary vector information, matrix information or tensor information with an identifier allotted beforehand to each element;

said distance evaluation function is a function having the vector information, matrix information or tensor information of a prescribed element configuration with an identifier allotted beforehand to each element; and

said distance calculating means calculates said distance information by reconfiguring each element of said arbitrary vector information, matrix information or tensor information such that the identifier of each element of said arbitrary vector information, matrix information or tensor information becomes the same as the identifier of the element of said prescribed element configuration, and inputting to said distance evaluation function.

25. An information classifying system, comprising:

an information classifying device, and an information terminal connected to said information classifying device through a communication line;

wherein

said information classifying device includes

population receiving unit configured to receive an arbitrary number of populations including pieces of sample information, from said information terminal,

distance calculating unit configured to calculate distance information between each of an arbitrary number of pieces of sample information included in said populations received by said population receiving unit and center of gravity of each of the arbitrary number of populations, using a distance evaluation function,

statistical information calculating unit configured to calculate statistical information using mean of sample distances and variance of sample distances with the mean of sample distances being the center, based on each group of distance information constituted by said distance information calculated by said distance calculating unit, associated with each said sample information belonging to each said population,

degree of assignment relevancy evaluating unit configured to evaluate degree of assignment relevancy, by calculating distance between the center of gravity of each said population and sample information as an object of classification using said distance calculating unit, and normalizing said distance using said statistical information calculated by said statistical information calculating unit on said calculated distance and performing a statistical test,

assignment destination determining unit configured to determine to which population said piece of information as the object of classification is to be assigned, according to the degree of assignment relevancy determined by said degree of assignment relevancy determining unit,

sample information classifying unit configured to assign and classify said piece of sample information as the object of classification to the population determined by said assignment destination determining unit, and

classified population passing unit configured to pass a classified population to which said piece of sample information as the object of classification has been assigned and classified by said sample information classifying unit to said information terminal; and

said information terminal includes

population passing unit configured to pass said arbitrary number of populations to said information classifying device, and

classified population receiving unit configured to receive said classified population from said information classifying device.

26. The information classifying system according to claim 25, wherein

said piece of sample information as the object of classification is an arbitrary vector information, matrix information or tensor information with an identifier allotted beforehand to each element;

said distance evaluation function is a function having the vector information, matrix information or tensor information of a prescribed element configuration with an identifier allotted beforehand to each element; and

said distance calculating unit is configured to calculate said distance information by reconfiguring each element of said arbitrary vector information, matrix information or tensor information such that the identifier of each element of said arbitrary vector information, matrix information or tensor information becomes the same as the identifier of the element of said prescribed element configuration, and inputting to said distance evaluation function.

27. An information classifying system comprising:

an information classifying device, and an information terminal connected to said information classifying device through a communication line;

wherein

said information classifying device includes

sample information receiving unit for receiving a piece of sample information as an object of classification from said information terminal,

distance calculating unit for calculating distance information between each of an arbitrary number of pieces of sample information including sample information as an object of classification received by said sample information receiving unit and center of gravity of each of the arbitrary number of populations including said sample information, using a distance evaluation function,

statistical information calculating unit for calculating statistical information, using mean of sample distances and variance of sample distances with the mean of sample distances being the center, based on each group of distance information constituted by said distance information calculated by said distance calculating unit, associated with each said sample information belonging to each said population,

degree of assignment relevancy evaluating unit for evaluating degree of assignment relevancy, by calculating distance between the center of gravity of each said population and sample information as an object of classification using said distance calculating unit, and normalizing said distance using said statistical information calculated by said statistical information calculating unit on said calculated distance and performing a statistical test,

assignment destination determining unit for determining to which population said piece of information as the object of classification is to be assigned, according to the degree of assignment relevancy determined by said degree of assignment relevancy determining unit, and

population identifying information passing unit for passing population identifying information identifying the population determined by said assignment destination determining unit to said information terminal; and

said information terminal includes

sample information passing unit for passing said piece of sample information as the object of classification to said information classifying device, and

population identifying information receiving unit for receiving said population identifying information from said information classifying device.

28. The information classifying system according to claim 27, wherein

said piece of sample information as the object of classification is an arbitrary vector information, matrix information or tensor information with an identifier allotted beforehand to each element;

said distance evaluation function is a function having the vector information, matrix information or tensor information of a prescribed element configuration with an identifier allotted beforehand to each element; and

said distance calculating unit calculates said distance information by reconfiguring each element of said arbitrary vector information, matrix information or tensor information such that the identifier of each element of said arbitrary vector information, matrix information or tensor information becomes the same as the identifier of the element of said prescribed element configuration, and inputting to said distance evaluation function.