# US 10,373,028 B2 · Pattern recognition device, pattern recognition method, and computer program product

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 10,373,028 B2
- Document ID: US10373028B2
- Status: Granted
- Kind: B2
- Application number: 15/708,367
- Filed: 2017-09-19
- Published: 2018-01-04
- Granted: 2019-08-06
- Earliest claimed: 2015-05-11
- Assignees: Kabushiki Kaisha Toshiba; Toshiba Digital Solutions Corporation
- Applicants: Kabushiki Kaisha Toshiba; Toshiba Digital Solutions Corporation
- Inventors: Soichiro Ono; Hiroyuki Mizutani
- Web page: https://patentgrep.com/patent/US10373028B2

## CPC

- G06V 10/267: {by performing operations on regions, e.g. growing, shrinking or watersheds}
- G06F 18/2193: {based on specific statistical tests}
- G06F 18/2415: based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- G06N 20/00: Machine learning
- G06V 10/424: Syntactic representation, e.g. by using alphabets or grammars
- G10L 25/03: characterised by the type of extracted parameters
- G10L 25/27: characterised by the analysis technique
- G10L 25/45: characterised by the type of analysis window
- G06V 30/287: {of Kanji, Hiragana or Katakana characters}
- G06V 30/293: {of characters other than Kanji, Hiragana or Katakana}
- G10L 15/14: using statistical models, e.g. Hidden Markov Models [HMMs] (G10L15/18 takes precedence)
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING

## Abstract

According to an embodiment, a pattern recognition device is configured to divide an input signal into a plurality of elements, convert the divided elements into feature vectors having the same dimensionality to generate a set of feature vectors, and evaluate the set of feature vectors using a recognition dictionary including models corresponding to respective classes, to output a recognition result representing a class or a set of classes to which the input signal belongs. The models each include sub-models each corresponding to one of possible division patterns in which a signal to be classified into a class corresponding to the model can be divided into a plurality of elements. A label expressing a model including a sub-model conforming to the set of feature vectors, or a set of labels expressing a set of models including sub-models conforming to the set of feature vectors is output as the recognition result.

## Claims

### Claim 1

A pattern recognition device comprising:
processing circuitry configured to function as:
a division unit configured to divide an input signal into a plurality of elements;
a feature extracting unit configured to convert the divided elements into feature vectors having the same dimensionality, and generate a set of feature vectors; and
a recognition unit configured to evaluate the set of feature vectors using a recognition dictionary, and output a recognition result representing a class or a set of classes to which the input signal belongs, wherein
the recognition dictionary includes models corresponding to respective classes,
the models each include sub-models each corresponding to one of possible division patterns in which a signal to be classified into a class corresponding to the model can be divided into a plurality of elements,
each sub-model has a state corresponding to each element divided based on a division pattern corresponding to the sub-model, the state being expressed by a function of labels representing a feature vector and the state, and
the recognition unit outputs, as the recognition result, a label expressing a model including a sub-model conforming to the set of feature vectors, or a set of labels expressing a set of models including sub-models conforming to the set of feature vectors.

### Claim 2 (depends on 1)

The device according to claim 1, wherein, in the sub-model, each state is linearly ordered, and transition from a higher state to a lower state is prohibited.

### Claim 3 (depends on 1)

The device according to claim 1, wherein
the recognition dictionary includes a reject model as a model not corresponding to any class, and
the recognition unit outputs information indicating that the recognition result is not obtained when the set of feature vectors conforms to the reject model.

### Claim 4 (depends on 1)

The device according to claim 1, wherein at least one of the models included in the recognition dictionary further has a noise state as a state not corresponding to any element included in a signal to be classified into a class corresponding to the model.

### Claim 5 (depends on 1)

The device according to claim 1, wherein, in a state of a sub-model, a function held by the state is determined based on a set of feature vectors to be an input to the state when a signal serving as learning data is input to the pattern recognition device, and a probability of input.

### Claim 6

A pattern recognition method executed by a pattern recognition device, comprising:
dividing an input signal into a plurality of elements by the pattern recognition device;
converting the divided elements into feature vectors having the same dimensionality, and generating a set of feature vectors by the pattern recognition device; and
evaluating the set of feature vectors using a recognition dictionary, and outputting a recognition result representing a class or a set of classes to which the input signal belongs by the pattern recognition device, wherein
the recognition dictionary includes models corresponding to respective classes,
the models each include sub-models each corresponding to one of possible division patterns in which a signal to be classified into a class corresponding to the model can be divided into a plurality of elements,
each sub-model has a state corresponding to each element divided based on a division pattern corresponding to the sub-model, the state being expressed by a function of labels representing a feature vector and the state, and
a label expressing a model including a sub-model conforming to the set of feature vectors, or a set of labels expressing a set of models including sub-models conforming to the set of feature vectors is output as the recognition result at the outputting.

### Claim 7

A computer program product comprising a non-transitory computer-readable medium including programmed instructions, the instructions causing a computer to implement:
a function of a division unit configured to divide an input signal into a plurality of elements;
a function of a feature extracting unit configured to convert the divided elements into feature vectors having the same dimensionality, and generate a set of feature vectors; and
a function of a recognition unit configured to evaluate the set of feature vectors using a recognition dictionary, and output a recognition result representing a class or a set of classes to which the input signal belongs, wherein
the recognition dictionary includes models corresponding to respective classes,
the models each include sub-models each corresponding to one of possible division patterns in which a signal to be classified into a class corresponding to the model can be divided into a plurality of elements,
each sub-model has a state corresponding to each element divided based on a division pattern corresponding to the sub-model, the state being expressed by a function of labels representing a feature vector and the state, and
the recognition unit outputs, as the recognition result, a label expressing a model including a sub-model conforming to the set of feature vectors, or a set of labels expressing a set of models including sub-models conforming to the set of feature vectors.
