FFT feature extraction on Unico

[SRodr.4]

Hello, I recently started working with Unico GUI software and I need to do a classification task for soundwaves with already gathered data using the decision tree from the MLC but I would like the classification based on features that can be extracted from the frequency domain (FFT) such as harmonics, spectral slope, pitch, etc. I am working in offline mode and first of all I would also need to compute the FFT of the loaded data but I haven't been able to do this in the Unico software. Is there a way to build the decision tree using features liek these or the only features available are the ones on the MLC (mean, variance, energy, peak to peak, etc.). Does anybody have a solution for this issue?

[Federica Bossi]

Hi @Sergio Rodriguez​ ,

Welcome to ST Community!

In the MLC tool of Unico GUI, we support only the features which are available in our MLC devices. Those features are documented in our MLC application notes.

Although there is not a specific FFT feature available on MLC, you can use the available filters and features to implement the desired classification, based on the available data logs.

Useful features for the type of classification described might be: Zero-crossing (with different thresholds) and Peak-to-peak.

Further improvements on the classification might be obtained by adding one or more filters (at different bandwidths) and computing the feature Energy on the filtered signals. 

If my reply answered your question, please click on Select as Best at the bottom of this post. This will help other users with the same issue to find the answer faster