Approach to sounds recognition with Machine or Deep Learning techniques?

Hi,

I'm working about a problem of sounds recognition: I could have ten different noises, and I need the computer can classify them, telling me which noise is when I supply as input a microphone signal or a registered audio file.

In other words, I would like to write a code with capabilities similar to the one described in this video, but using sounds instead of images:

https://it.mathworks.com/videos/machine-learning-made-easy-106181.html

Could someone help me in terms of approach to the problem? Which file format I need to train the network (mp3, wav...)? Which neural network could I use? Could you suggest me some references and examples?

Thank you so much, kind regards.

Answer 1

[Krishna]

Hi,

To classify different sound using MATLAB you can look at the following example,

https://in.mathworks.com/help/audio/gs/classify-sound-using-deep-learning.html

This example uses Recurrent Neural Networks (RNN) for classification of audio Signals particularly Long Short-Term Memory (LSTM) network. These networks are the build for time series analysis and favourite choice by researchers for audio signal analysis.

Apart from theses if you want to use less parameters for training you can also use convolutional neural networks for training audio time series signals particularly 1-D convolution network.

Look at the following examples to learn more,

https://in.mathworks.com/help/deeplearning/ug/sequence-classification-using-1-d-convolutions.html

The data needs to be stored in form of datastores for deep learning analysis on audio signals. Look at this MATLAB answer for more information,

https://in.mathworks.com/matlabcentral/answers/501364-classify-sound-using-deep-learning-database-how?s_tid=answers_rc1-3_p3_Topic

Hope this helps.