Recommending measurements based on detected waveform type

Patent metadata

Field	Value
Publication number	US-10585121-B2
Application number	US-201615262406-A
Country	US
Kind code	B2
Filing date	Sep 12, 2016
Priority date	Sep 12, 2016
Publication date	Mar 10, 2020
Grant date	Mar 10, 2020

Abstract

An oscilloscope including an input port for receiving training data including waveforms and corresponding known classifications and a processor for training a plurality of classifiers on the training data. Training includes iteratively applying each classifier to each waveform of the training data to obtain corresponding predicted waveform classifications and comparing the predicted waveform classifications with the known classifications. Classifiers are corrected when predicted waveform classifications does not match the known classifications. Models for each classification are constructed with suggested measurements or actions. Subsequently, live waveform data is captured by the oscilloscope and the classifiers are applied to the live data. When a confidence value for a single classification exceeds a threshold, the waveform data is classified, and suggested measurements or actions are implemented in the oscilloscope based on the classification.

First claim

We claim:

1. A method for training a test and measurement instrument to suggest measurements, the method comprising:
receiving training data including waveforms and corresponding known classifications via an input port; and
training, via a processor, a plurality of classifiers on the training data by:
iteratively applying each classifier to each waveform of the training data to obtain corresponding predicted waveform classifications;
comparing the predicted waveform classifications with the known classifications; and
correcting each classifier that outputs a predicted waveform classification that does not match the known classification for the corresponding waveform.

2. The method of claim 1 , wherein the classifiers include a decision forest, and wherein training the classifiers involves growing decision trees in the decision forest.

3. The method of claim 2 , further comprising storing a confidence threshold to each classifier, the confidence threshold indicating a minimum percentage of decision trees in the decision forest that output a positive classification to result in a positive classification for the corresponding classifier.

4. The method of claim 1 , further comprising employing cross validation of the training data of waveforms to further correct the classifiers.

5. The method of claim 1 , further comprising employing feature selection on each classifier to reduce a corresponding set of classifier inputs by removing inputs that are not useful for the corresponding classification.

6. The method of claim 5 , further comprising storing a set of one or more suggested actions to each model, the suggested actions including instructions executed by the test and measurement system upon matching measurement data with a classification of a classifier corresponding to the model.

7. The method of claim 1 , further comprising:
receiving measurement data corresponding to a measured waveform or bus input;
applying each of the classifiers to the measurement data to determine a classification for the measured waveform;
selecting a model corresponding to the classifier returning the determined classification, the model corresponding to one or more suggested actions;
receiving, at a user control, a selection from a user; and
configuring, in response to the selection from the user, the test and measurement instrument to employ the suggested actions corresponding to the selected model when capturing the measured waveform.

8. The method of claim 7 , wherein determining the classification for the measured waveform includes producing a confidence value for each of the classifiers, and wherein the classification is determined when a single classifier produces a confidence value in excess of a stored confidence threshold.

9. A non-transitory computer-readable medium comprising a set of instructions which instructions, when executed by a processor of a test and measurement instrument, cause the test and measurement instrument to:
receive training data including waveforms and corresponding known classifications via an input port; and
train, via the processor, a plurality of classifiers on the training data by:
iteratively applying each classifier to each waveform of the training data to obtain corresponding predicted waveform classifications;
comparing the predicted waveform classifications with the known classifications; and
correcting each classifier that outputs a predicted waveform classification that does not match the known classification for the corresponding waveform.

10. The non-transitory computer-readable medium of claim 9 , further comprising a second set of instructions, which second set of instructions, when executed by the processor, cause the test and measurement instrument to:
receive measurement data corresponding to a measured waveform or bus input;
apply each of the classifiers to the measurement data to determine a classification for the measured waveform;
select a model corresponding to the classifier returning the determined classification, the model corresponding to one or more suggested actions;
receive, at a user control, a selection from a user; and
configure, in response to the selection from the user, the test and measurement instrument to employ the suggested actions corresponding to the selected model when capturing the measured waveform.

Assignees

• Tektronix Inc

Inventors

• Absher Ian R
• Strong Kraig M

Classifications

• G01R13/22 Primary

  Circuits therefor · CPC title

• G06N20/00

  Machine learning · CPC title

• G01R13/02

  for displaying measured electric variables in digital form · CPC title

• G01R13/00

  Arrangements for displaying electric variables or waveforms · CPC title

• G01R23/16

  Spectrum analysis; Fourier analysis · CPC title