Week 6 Starting Point

This week I'm going to try to find a Neural Network code I can use to input audio into. I think I'm going to see what's been previously done and studied to figure out what aspect of audio I should be testing, i.e. pitch, spectrograms, Log-Mel, etc.

This link is the TensorFlow audio recognition materials on github again. This is the tutorial page itself! Might want to start with this since it's been previously mentioned and already has audio samples to work with even if they are not speech samples. Could be good to just get used to running an NN and seeing how much time it takes.

This is a github featuring NN code for detecting music. This may be easy to edit for my purposes in speech recognition and also testing what data to test in an audio file.

This is a paper specifically about using an NN for speech recognition. Looks like it's a heavy read, but might be helpful in deciding weights or finding errors others have run into. Uses a convolutional NN (CNN).
This is another paper that uses CNNs.

I also found a book that has a chapter on Neural Networks while coding in the Programming language. That might be helpful considering I'm currently taking a Programming class and it's a little easier to work with than Python, but I'll have to read more of the chapter to see if there are obvious limitations and if it can take audio data.

I think testing the waters in audio recognition is the first step to go. I don't think I'll worry about rigging or anything this week, just pure NN research and how to go about it.

Comments