Sound Generation in Python: Turning Your Data into Music

Not long ago, I published an article entitled “The Sound that Data Makes”. The goal was turning data — random noise in this case — into music. The hope was that by “listening” to your data, you could gain a different kind of insights, not conveyed by visualizations or tabular summaries.

This article is a deeper dive on the subject. First, I illustrate how a turn the Riemann zeta function, at the core of the Riemann conjecture, into music. It constitutes an introduction to scientific programming, the MPmath library, and complex number algebra in Python. Of course this works with any other math function. Then I explain how to use the method on real data, create both a data sound track and a data video, and combine both. This material and the accompanying PDF document corresponds to one of the projects offered to participants in the AI/ML technical training and certification designed by my research lab. More about this program, here.

Benefits of Data Sonification

Data visualizations offer colors and shapes, allowing you to summarize multiple dimensions in one picture. Data animations (videos) go one step further, adding a time dimension. You can find many on my YouTube channel. Then, sound adds multiple dimensions: amplitude, volume and frequency over time. Producing pleasant sound, with each musical note representing a multivariate data point, is equivalent to data binning or bukectization.

Stereo and the use of multiple musical instruments (synthesized) add more dimensions. Once you have a large database of data music, you can use it for generative AI: sound generation to mimic existing datasets. Of course, musical AI art is another application, all the way to creating synthetic movies.

To read the full article with documented source code, follow this link. No sign-up required.