Matlab - autoencoder for speech signal

Asked Aug 20 '16 at 15:03

Active Aug 20 '16 at 15:22

Viewed 149 times

I want to use the trainAutoencoder function from matlab to find the 30 main patterns of 300 speech signals.

I tried to use this function, and use plotWeigths to see the patterns (weigths) but it seems that this is only for pictures and not for speech signals.

I want to get 30 graph that each one length is the same as my speech signals and I can get any of my 300 signals as a combination of the 30 main patterns with minimum error.

This is what I tried:

 autoenc2 = trainAutoencoder(cells,30   ,'MaxEpochs',400, ...
    'EncoderTransferFunction','satlin',...
    'DecoderTransferFunction','purelin',...
    'L2WeightRegularization',0.01,...
    'SparsityRegularization',4,...
    'SparsityProportion',0.10);

plotWeights(autoenc2);

cells = is a cell array of size 1x300, each cell contain one of the 300 signals.

Somebody can help me. Thanks,

edited Aug 20 '16 at 15:22

asked Aug 20 '16 at 15:03

brianch

1

And what is the question? – NKN Aug 20 '16 at 15:12
How can I do it? Which function can I use? – brianch Aug 20 '16 at 15:13
You should add what you have tried so far, and ask specific questions on the code you have. SO does not provide a whole tutorial for your question. – NKN Aug 20 '16 at 15:16
edit the original question – brianch Aug 20 '16 at 15:23

Matlab - autoencoder for speech signal

0 Answers0