This repository is a representation of my work for my Belmont AET Graduate program thesis. In this GitHub project I will show you the path of learning I took for ML and how I learned to apply it to audio!
For this project we will be using Python as our programming language. Going into this project I had no experience with Machine Learning and Deep Learning so I was starting from scratch. If you have experience with these techniques feel free to skip this section!
For learning the basics I followed this course on Udemy! Usually it is on sale, but there are plenty of free courses you can find. I just found this one the most useful for my own learning! Machine Learning Bootcamp
This section of the project is in two parts: Machine Learning Algos and Neural Networks. The Machine Learning Algorithms was me getting familiar with the basic concpets and implementing them based on the course listed above. For the second section, we build off the concepts from Machine Learnign Algorithms and start to using Neural Networks!
ML concepts you should know before heading onto the audio:
- Loss functions & activation functions
- Deep Neural Networks
- Supervised Learning vs Unsupervised Learning
- Using Keras and SKLearn
- LSTMs
- Convolution
- Classification using NN
I was already quite familiar with digital audio concepts, but I found this great YouTube series for review of concepts! Sound of AI
Even though this resource uses MATLAB and not Python, it is an excellent resource if you are just getting started in the world of digital audio! Hack Audio
Audio concepts you should know before combining all that we learned!
- Intensity
- Frequency & Harmonics
- Dynamics & Decibels
- Timbre
- Sampling Rate & Bit Depth
- Mel Frequency Cepstrum Coefficients
- Spectrograms
- Fourier Transform
For this I followed along the Sound of AI playlist above to work on Neural Networks! This part of the project was mostly focused on genre classifcation and audio data preparation using the GTZAN data set.