Feature extraction from a signal and classification

I was wondering if anyone could help me with a few steps or even code to get started on feature extraction from a signal. I would like to extract the features of a signal and then classify them in the classification learner app.

The data I have is a vibration signal that varies between a set of healthy bearings and bearings that are faulty.

The aim is to extract the features and then compare them in the classification app with each other to identify when bearings are faulty.

Any help is appreciated!

Thanks

[Image Analyst]

Care to show us a good and bad signal? Or attach them if you want anyone to try anything with them. Perhaps you can call pwelch() and check for anomalies in the spectrum.

[Harry]

I uploaded the signals and the code.

Thank you

[Liyifei LiYiFei]

你好，我也有和你相似的问题，你解决了吗，如何对信号进行特征提取并进行故障分类

Answer 1

[Abel Babu]

Hi Harry,

This MATLAB example would be a good starting point:

http://in.mathworks.com/examples/signal/mw/signal-ex86966978-extracting-classification-features-from-physiological-signals

The example deals with the classification of physiological signals but the features used here can be applied to classification of signals in general. The extracted features can then be fed as features for the classification app.

As suggested by Image Analyst, spectral analysis can be used to generate more features as well.

Abel

[John BG]

bearings: don't seem quite same domain as phyisio, do they?

John BG

<mailto:[e-mail] [e-mail]>

[Jan]

+1. For Matlab all signals are a list of numbers only and it does not matter what the signal means physically.

[Aditya Baru]

Here's an example for feature extraction for bearing signals :) https://www.mathworks.com/help/predmaint/examples/wind-turbine-high-speed-bearing-prognosis.html https://www.mathworks.com/help/predmaint/examples/Rolling-Element-Bearing-Fault-Diagnosis.html

The new Predictive Maintenance Toolbox has a lot of capabilities that support this kind of workflow, so you could check out its documentation for more information.

Answer 2

[Zhao Lu]

MATLAB has EMD function, for EEMD, the code can be downloaded from

http://perso.ens-lyon.fr/patrick.flandrin/emd.html

[Francisco Navarro]

thanks!

Answer 3

[Rahmawati Rahmawati]

halo everyone, I am rahma and i am totally newbie in EEG data analysis. I got an assignment to make a classification between two conditions using spectral powers based on Raw EEG data which has been given by my Professor. but to be honest i don't know how to start with this. and here are the state: Sampling rate: 512 HZ Channel position: POz, PO1, PO2, PO3, PO4, Oz, O1, O2

any help and hints are totally helpful for me.

Thanks in advance.

[Smith Khare]

One need complete dataset to process and understand, then after the main work starts

Answer 4

[karthikeyan chandrasekar]

hi everyone can anyone tell me how to extract features using PCA for a signal ,i.e 8190x2 signal which is in text[matrix] format.

Thankyou in advance

[John Navarro]

By PCA did you mean Principal component analysis?

If so, PCA does not extract features, it evaluates their correlation and indicates the more useful ones. PCA is employed for feature selection, no feature extraction. It should be done according the expertise, the case of study, and the features of interest.

Answer 5

[Zhao Lu]

Use EMD or EEMD or CEEMDAN

[Francisco Navarro]

Hi! What function or toolbox do you recommend to use EEMD or NA-MEMD or CEEMDAN? Thank you!

[John Navarro]

Signal Processing. Neither of these options are in MATLAB, as far as I know.

Maybe in a forum as function made by other users.

EMD and VMD are the most similar functions you would find in the program.

[Zhao Lu]

https://www.mathworks.com/help/signal/ref/emd.html

http://bioingenieria.edu.ar/grupos/ldnlys/metorres/re_inter.htm#Codigos

Answer 6

[litha Mbangeni]

Can I simulate CEEMDAN in matlab

[Smith Khare]

Yes... There is a toolbox available. You can use it to analyze the signal

[Smith Khare]

https://github.com/ron1818/PhD_code/blob/master/EMD_EEMD/ceemdan.m

The above link could be of interest to you

Answer 7

[Shrey Joshi]

you can start by using features provided in feature extraction mode of signal labeler app.

https://www.mathworks.com/help/signal/ug/extract-signal-features.html