Has anyone attempted Principal component analysis on time series and acceleration data(ie. data from accelerometer and sensors) and tried compressing it as well as regenerating back the data with minimum error?
I wanted to know if this is possible using PCA.
If processing a data stream from a sensor, I would go for a method that you can implement recursively. You wouldn't want to be computing the eigenvalues of a large matrix, every time you collect a new sample.
It depends on the nature of the signal and the noise/interference. If it was a low-frequency signal, I would fit polynomials to the data stream. Otherwise, I would do a Fourier analysis. Both can be implemented recursively. So every time you get a sample in, you get an estimate out, with just a few operations. Of course, if you want the estimate to be reasonable, you will need to delay your estimate.
- Badges
- Science topic