Mapper component for a neuro-linguistic behavior recognition system

• US 10,373,062 B2
• Filed: 12/12/2014
• Issued: 08/06/2019
• Est. Priority Date: 12/12/2014
• Status: Active Grant

• Current Assignee: Intellective Ai, Inc.
• Original Assignee: Omni AI, Inc.
• Inventors: Seow, Ming-Jung, Xu, Gang, Yang, Tao, Cobb, Wesley Kenneth
• Primary Examiner(s): Lu, Kuen S
• Application Number:
  US14/569,034
• Publication Number:
  US 20160171096A1
• Time in Patent Office: 1,698 Days
• Field of Search: 707738
• CPC Class Codes: G06F 18/2137 based on criteria of topology preservation, e.g. multidimensional scaling or self-organising maps
• CPC Class Codes: G06F 18/23 Clustering techniques
• CPC Class Codes: G06F 18/28 Determining representative reference patterns, e.g. by averaging or distorting; Generating dictionaries
• CPC Class Codes: G06N 3/0409 Adaptive resonance theory [ART] networks
• CPC Class Codes: G06N 3/042 Knowledge-based neural networks; Logical representations of neural networks
• CPC Class Codes: G06N 3/047 Probabilistic or stochastic networks
• CPC Class Codes: G06N 3/088 Non-supervised learning, e.g. competitive learning
• CPC Class Codes: G06N 5/022 Knowledge engineering; Knowledge acquisition
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks
• CPC Class Codes: G06V 10/32 Normalisation of the pattern dimensions
• CPC Class Codes: G06V 10/762 using clustering, e.g. of similar faces in social networks
• CPC Class Codes: G06V 30/19127 Extracting features by transforming the feature space, e.g. multidimensional scaling; Mappings, e.g. subspace methods
• CPC Class Codes: G06V 30/1914 Determining representative reference patterns, e.g. averaging or distorting patterns; Generating dictionaries, e.g. user dictionaries
• CPC Class Codes: G06V 30/268 Lexical context
• CPC Class Codes: H01B 1/02 mainly consisting of metals or alloys

Abstract

Techniques are disclosed for generating a sequence of symbols based on input data for a neuro-linguistic model. The model may be used by a behavior recognition system to analyze the input data. A mapper component of a neuro-linguistic module in the behavior recognition system receives one or more normalized vectors generated from the input data. The mapper component generates one or more clusters based on a statistical distribution of the normalized vectors. The mapper component evaluates statistics and identifies statistically relevant clusters. The mapper component assigns a distinct symbol to each of the identified clusters.

Claims

1. A computer-implemented method, comprising:

receiving a first normalized vector of at least one feature value generated from first input data, the at least one feature value being associated with a first feature of a plurality of features, the plurality of features characterizing appearance and kinematic aspects of at least one object depicted in the first input data;

for the at least one feature value in the first normalized vector;

evaluating a distribution of at least one cluster in a cluster space corresponding to the first feature associated with the at least one feature value, and

mapping the at least one feature value to the at least one cluster based on the distribution;

updating the distribution of the at least one cluster based on the mapping;

determining, based on the updated distribution, whether or not to merge the at least one cluster with a further cluster from the cluster space; and

outputting, to a lexical analyzer, at least one symbol associated with the at least one cluster, the lexical analyzer configured to build a dictionary based on the outputted at least one symbol.

2. The computer-implemented method of claim 1, further comprising:

receiving a second normalized vector of the at least one feature value generated from second input data; and

for the at least one feature value in the second normalized vector;

mapping the at least one feature value to the at least one cluster,

updating the distribution of the at least one cluster based on the mapping, and

upon determining that the at least one cluster has statistical significance based on the updated distribution, outputting, to the lexical analyzer, the at least one symbol associated with the at least one cluster.

3. The computer-implemented method of claim 1, further comprising, upon determining that the at least one cluster and at least another cluster overlap in the cluster space based on the evaluated distribution of the at least one cluster and the at least other cluster, merging the at least one cluster and the at least other cluster.

4. The computer-implemented method of claim 1, wherein a statistical significance of the at least one cluster is determined based on a statistical significance score that indicates a specified amount of the at least one feature value mapping to the at least one cluster over time.

5. The computer-implemented method of claim 4, further comprising, upon determining that the at least one feature value mapping to the at least one cluster has not been received after a period based on a function of time, decaying the statistical significance of the at least one cluster.

6. The computer-implemented method of claim 1, wherein the distribution of the at least one cluster includes a mean, variance, and standard deviation.

7. The computer-implemented method of claim 1, wherein each value in the first normalized vector is within a range of 0 and 1, inclusive.

8. The method of claim 1, further comprising sending an unusualness score to the lexical analyzer.

9. A non-transitory computer-readable storage medium comprising:

computer-executable instructions that, when executed on at least one processor, cause the at least one processor to;

receive a first normalized vector of at least one feature value generated from first input data, the at least one feature value being associated with a first feature of a plurality of features, the plurality of features characterizing appearance and kinematic aspects of at least one object depicted in the first input data;

for the at least one feature value in the first normalized vector;

evaluate a distribution of at least one cluster in a cluster space corresponding to the first feature associated with the at least one feature value, and

map the at least one feature value to the at least one cluster based on the distribution;

update the distribution of the at least one cluster based on the mapping;

determine, based on the updated distribution, whether or not to merge the at least one cluster with a further cluster from the cluster space; and

output, to a lexical analyzer, at least one symbol associated with the at least one cluster, the lexical analyzer configured to build a dictionary based on the outputted at least one symbol.

10. The non-transitory computer-readable storage medium of claim 9, wherein the computer-executable instructions further comprise instructions that cause the at least one processor to:

receive a second normalized vector of the at least one feature value generated from second input data; and

for the at least one feature value in the second normalized vector;

map the at least one feature value to the at least one cluster,

update the distribution of the at least one cluster based on the mapping, and

upon determining that the at least one cluster has statistical significance based on the updated distribution, output, to the lexical analyzer, the at least one symbol associated with the at least one cluster.

11. The non-transitory computer-readable storage medium of claim 9, wherein determining whether to merge the at least one cluster with a further cluster from the cluster space is based on whether there is overlap in the cluster space between the at least one cluster and the further cluster.

12. The non-transitory computer-readable storage medium of claim 9, wherein a statistical significance of the at least one cluster is determined based on a statistical significance score that indicates a specified amount of the at least one feature value mapping to the at least one cluster over time.

13. The non-transitory computer-readable storage medium of claim 12, wherein computer-executable instructions further comprise instructions that cause the at least one processor to:

upon determining that the at least one feature value mapping to the at least one cluster has not been received after a period based on a function of time, decay the statistical significance of the at least one cluster.

14. The non-transitory computer-readable storage medium of claim 9, wherein the distribution of the at least one cluster includes a mean, variance, and standard deviation.

15. The non-transitory computer-readable storage medium of claim 9, wherein each value in first normalized vector is within a range of 0 and 1, inclusive.

16. A system, comprising:

a processor; and

a memory in communication with the processor and storing code, that, when executed on the processor, causes the processor to;

receive a first normalized vector of at least one feature value generated from first input data, the at least one feature value being associated with a first feature of a plurality of features, the plurality of features characterizing appearance and kinematic aspects of at least one object depicted in the first input data;

for the at least one feature value in the first normalized vector;

evaluate a distribution of at least one cluster in a cluster space corresponding to the first feature associated with the at least one feature value, and

map the at least one feature value to the at least one cluster based on the distribution;

update the distribution of the at least one cluster based on the mapping;

determine, based on the updated distribution, whether or not to merge the at least one cluster with a further cluster from the cluster space; and

output, to a lexical analyzer, at least one symbol associated with the at least one cluster, the lexical analyzer configured to build a dictionary based on the outputted at least one symbol.

17. The system of claim 16, wherein the code, when executed, further causes the processor to:

receive a second normalized vector of the at least feature value generated from second input data; and

for the at least one feature value in the second normalized vector;

map the at least one feature value to the at least one cluster,

update the distribution of the at least one cluster based on the mapping, and

upon determining that the at least one cluster has statistical significance based on the updated distribution, output, to the lexical analyzer, the at least one symbol associated with the at least one cluster.

18. The system of claim 16, wherein the determining whether to merge the at least one cluster with the further cluster from the cluster space is based on whether there is overlap in the cluster space between the at least one cluster and the further.

19. The system of claim 16, wherein a statistical significance of the at least one cluster is determined based on a statistical significance score that indicates a specified amount of the at least one feature value mapping to the at least one cluster over time.

20. The system of claim 19, wherein the code, when executed, further causes the processor to:

upon determining that the at least one feature value mapping to the at least one cluster has not been received after a period based on a function of time, decay the statistical significance of the at least one cluster.

21. The system of claim 16, wherein each value in first normalized vector is within a range of 0 and 1, inclusive.