How can I train CNN(Matlab Classification Learner) with spectogram of a signal?

Hi there,

I have a dataset consist of IQ datas. I use "spectogram(x)" command to take spectograms of an signal. I want to train those dataset with CNN in MATLAB Classification Learner. However, Classification Learner not accept an array in table(My IQ data's are in size of 1290x1 and I have 66560 data). My question is how can I train CNN with those spectograms?

Detailed information about my dataset;

"Dataset 66560x1 table"

"IQ data: 1280x1 complex double"

Thank you.

Answer 1

[Drew]

As of 23b, the Classification Learner app, which is part of the Statistics and Machine Learning Toolbox (SMLT), is not designed to train convolutional neural networks (CNN). The Classification Learner app supports ordinary tabular datasets with one predictor in each column and one obversation in each row, and can train certain neural networks supported by the SMLT function "fitcnet".

To train a convolutional neural network (CNN), use functionality from the Deep Learning Toolbox (DLT). For this, see doc such as the following, or other related doc pages.

• Simple example using command line: https://www.mathworks.com/help/deeplearning/ug/create-simple-deep-learning-network-for-classification.html
• Deep Network Designer app: https://www.mathworks.com/help/deeplearning/ug/train-networks-using-deep-network-designer.html

If this answer helps you, please remember to accept the answer.