Granted Patent B2

US 12,299,071 · App. 17/447,012 · Granted May 13, 2025

Inventors: Akihiro Yamaguchi (Kita Tokyo, JP); Ken Ueno (Tachikawa Tokyo, JP)

Assignee: Kabushiki Kaisha Toshiba

G06F18/214G06F2218/02G06F2218/10G06F2218/20

Abstract

Claims (84)

1. An information processing apparatus comprising:

an adjuster configured to adjust a shape of a reference waveform pattern corresponding to time series data and used to classify the time series data, on a basis of a value of a factor parameter corresponding to the time series data; and

a feature calculator configured to calculate a feature of a waveform of the time series data on a basis of the shape of the reference waveform pattern after the adjustment,

wherein

the reference waveform pattern before the adjustment is a shape that does not correspond to the value of the factor parameter, but the reference waveform pattern after the adjustment is a shape that corresponds to the value of the factor parameter.

2. The information processing apparatus according to claim 1 , further comprising:

a classification device configured to acquire a classification result for the time series data by inputting the feature into a classifier.

3. The information processing apparatus according to claim 1 , wherein

if the value of the factor parameter varies, the waveform of the time series data deforms,

the feature is based on a distance between the reference waveform pattern and the waveform of the time series data, and

the feature based on the reference waveform pattern after the adjustment is smaller than the feature based on the reference waveform pattern before the adjustment.

4. The information processing apparatus according to claim 1 , further comprising:

an input device; and

an output device, wherein

the input device receives a specific value with respect to the factor parameter,

the adjuster executes the adjustment of the shape of the reference waveform pattern on a basis of the specific value, and

the output device outputs information related to the adjustment.

5. The information processing apparatus according to claim 1 , further comprising:

an input device; and

an output device, wherein

the input device receives a transition range or a plurality of specific values of the value of the factor parameter,

the adjuster executes the adjustment of the shape of the reference waveform pattern on a basis of a plurality of values included in the transition range or the plurality of specific values, and

the output device outputs information related to each position of a plot forming the reference waveform pattern after executing the adjustment.

6. An information processing apparatus comprising:

an adjuster configured to adjust a shape of a reference waveform pattern corresponding to time series data and used to classify the time series data, on a basis of a value of a factor parameter corresponding to the time series data; and

a feature calculator configured to calculate a feature of a waveform of the time series data on a basis of the shape of the reference waveform pattern after the adjustment,

wherein

the adjuster executes the adjustment on a basis of an output of an estimation model obtained by inputting the value of the factor parameter into the estimation model, and

the estimation model estimates a deformation amount of the shape of the reference waveform pattern before and after the adjustment.

7. The information processing apparatus according to claim 6 , wherein

the deformation amount is expressed as a distance that each plot forming the reference waveform pattern has moved due to the adjustment.

8. An information processing apparatus comprising:

an adjuster configured to adjust a shape of a reference waveform pattern corresponding to time series data and used to classify the time series data, on a basis of a value of a factor parameter corresponding to the time series data;

a feature calculator configured to calculate a feature of a waveform of the time series data on a basis of the shape of the reference waveform pattern after the adjustment;

a classification device configured to acquire a classification result for the time series data by inputting the feature into a classifier;

a classifier updater; and

a reference waveform pattern updater,

wherein

the feature calculator calculates a training feature related to training time series data in which a correct class into which the data should be classified is indicated, on a basis of the training time series data and a training reference waveform pattern corresponding to the training time series data,

the classification device acquires a training classification result for the training time series data by inputting the training feature into the classifier,

the classifier updater updates a value of a parameter of the classifier on a basis of the training classification result and the correct class, and

the reference waveform pattern updater updates the training reference waveform pattern on a basis of the training classification result and the correct class, and uses the updated training reference waveform pattern to generate the reference waveform pattern.

9. The information processing apparatus according to claim 6 , further comprising:

an input device; and

an output device, wherein

the input device receives a specification related to the estimation model,

the adjuster executes the adjustment of the shape of the reference waveform pattern on a basis of the specified estimation model, and

the output device outputs information related to the adjustment.

10. The information processing apparatus according to claim 8 , wherein

the updating performed by the classifier updater and the reference waveform pattern updater is based on gradient descent.

11. An information processing apparatus comprising:

an adjuster configured to adjust a shape of a reference waveform pattern corresponding to time series data and used to classify the time series data, on a basis of a value of a factor parameter corresponding to the time series data;

a feature calculator configured to calculate a feature of a waveform of the time series data on a basis of the shape of the reference waveform pattern after the adjustment;

a classification device configured to acquire a classification result for the time series data by inputting the feature into a classifier;

a classifier updater;

a reference waveform pattern updater; and

an estimation model updater,

wherein

the feature calculator calculates a training feature related to training time series data in which a correct class into which the data should be classified is indicated, on a basis of the training time series data and a training reference waveform pattern corresponding to the training time series data,

the classification device acquires a training classification result for the training time series data by inputting the training feature into the classifier,

the classifier updater updates a value of a parameter of the classifier on a basis of the training classification result and the correct class, and

the reference waveform pattern updater updates the training reference waveform pattern on a basis of the training classification result and the correct class, and uses the updated training reference waveform pattern to generate the reference waveform pattern,

the adjuster adjusts a shape of the training reference waveform pattern used to calculate the training feature on a basis of an output of an estimation model obtained by inputting the value of the factor parameter corresponding to the training time series data into the estimation model, the estimation model estimating a deformation amount of the shape of the reference waveform pattern before and after the adjustment, and

the estimation model updater updates a parameter of the estimation model such that the shape of the adjusted training reference waveform pattern approaches the shape of the updated training reference waveform pattern.

12. An information processing method comprising:

adjusting a shape of a reference waveform pattern corresponding to time series data and used to classify the time series data, on a basis of a value of a factor parameter corresponding to the time series data; and

calculating a feature of a waveform of the time series data on a basis of the shape of the reference waveform pattern after the adjustment,

wherein

the reference waveform pattern before the adjustment is a shape that does not correspond to the value of the factor parameter, but the reference waveform pattern after the adjustment is a shape that corresponds to the value of the factor parameter.

13. A non-transitory storage medium storing a program executed by a computer, the program comprising:

adjusting a shape of a reference waveform pattern corresponding to time series data and used to classify the time series data, on a basis of a value of a factor parameter corresponding to the time series data; and

calculating a feature of a waveform of the time series data on a basis of the shape of the reference waveform pattern after the adjustment,

wherein

the reference waveform pattern before the adjustment is a shape that does not correspond to the value of the factor parameter, but the reference waveform pattern after the adjustment is a shape that corresponds to the value of the factor parameter.

14. An information processing method comprising:

adjusting a shape of a reference waveform pattern corresponding to time series data and used to classify the time series data, on a basis of a value of a factor parameter corresponding to the time series data;

calculating a feature of a waveform of the time series data on a basis of the shape of the reference waveform pattern after the adjustment; and

executing the adjustment on a basis of an output of an estimation model obtained by inputting the value of the factor parameter into the estimation model,

wherein the estimation model estimates a deformation amount of the shape of the reference waveform pattern before and after the adjustment.

15. A non-transitory storage medium storing a program executed by a computer, the program comprising:

adjusting a shape of a reference waveform pattern corresponding to time series data and used to classify the time series data, on a basis of a value of a factor parameter corresponding to the time series data;

calculating a feature of a waveform of the time series data on a basis of the shape of the reference waveform pattern after the adjustment; and

executing the adjustment on a basis of an output of an estimation model obtained by inputting the value of the factor parameter into the estimation model,

wherein the estimation model estimates a deformation amount of the shape of the reference waveform pattern before and after the adjustment.

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (18)