Hi. I would like to ask about my project that anyone know which processing block should I use for data collected as current values through a piezoelectric transducer. and use this to predict continuous motions?
Hello @martin449,
What is the sampling frequency you want to use?
If the sampling frequency is high enough (let’s say above 60Hz), you can use the spectral feature processing block. Alternatively you can also use the raw data processing block and / or the flatten processing block.
Hi louis, the sampling frequency is 100Hz. But when I use the spectral feature processing block, the calculated spectral features are something like this. Is this because the collected values are too small? and the accuracy for the model is only below 6%.
Hello @martin449,
I’d try with the raw data first to see if you have any good results and then maybe add a flatten in addition to the raw data.
Do you have an example of what your data samples look like?
Regards,
Louis
Hi Louis,
My data sample looks something like this one (This is data collected when running).
Kind regards,
Martin
Hi @louis, I just want to ask about the Raw data processing block. When generating features for the neuron network input, is each sample data processed as a feature? Or how the final features processed? As my data has more than 12000 samples (training windows) and the number of input feature is 300. Thanks.