Music Analysis via Python

This is my passion project that will be in continuous development. So far I have created an algorithm that takes in microphone data, creates a wave file, creates an audio data profile consisting of the frequency values of a Fourier transform, and the MEL spectrogram of the audio. It then compares this profile with a MIDI database that matches the frequency values with their corresponding musical note values and returns the notes value.

A side note, this process of converting analog audio inputs into unique interpretations that can be used for data analysis, like Mel Spectrograms, is the same process used on submarines to detect and classify anomalies picked up by sonar.

My goal with this is to create a tool that helps ease the transition from practical guitar inputs to music theory concepts. It is a well-known meme within the guitar community that we are not the most attune to music theory concepts. It is much easier to naturally play guitar than it is to understand the underlying theoretical foundations that help describe the music you're playing.  My guitar teacher studied music theory at the University of Southern California and has always implied it is extremely useful to understand what you're playing and why it sounds the way it does. However, he tends to teach guitar from a practical standpoint rather than a heavily theoretical one. I took a music theory class in college to satisfy my art requirement and have been trying to deepen my understanding for years now.

Even the small progress I have made with this program has greatly contributed to my understanding of these fundamental music concepts.

I am currently working on a way to graphically display the notes and their relationships to each other using descriptions like intervals and scale degrees.

I believe that machine learning (ML) would be a great tool for this. I have been researching how to implement TensorFlow into an application like this. I am working on creating a database of .wav files labeled by their note. I believe that training a network to identify the MEL spectrogram of this audio data is a perfect application of ML.