Training a audio classification neural network with .csv's containing features over time

Hello,

I am trying to train a neural network to classify an audio signal, and I am using MFCCs and a variety of other features that vary over time. My work so far has left me with two subfolders for each class 'Positive Diagnosis' and 'Negative Diagnosis' containing csv files whos first row contains feature names, and later rows contain the extracted feature values over time.

Feature 1 | Feature 2 | ...

val1(t1) val2(t1)

val1(t2) val2(t2)

...

I have found plenty of ways to create datastore objects for audio and image files that seem to integrate well into Matlabs Neural Network workflow, but the tabularTextDatastore doesnt seem to support labels and I'm at a loss at how to actually feed data of this format into training with appropriate labels. Is it possible to do feature extraction like this on my own before neural network training, and if so, how do I use this data?

Answer 1

[Swetha Polemoni]

Hi

It is my understanding that you want to feed the network with tabularTextDatastore. You can use transform to convert the tabularTextDatastore to other formats that can be easily fed to Matlab's neural networks.