AudioCompression using AutoEncoder

Machine Learning is replacing previous techniques in each field

So I thought why not try Machine Learning to compress the audio

Nodes for each layer can be tweeked, you can compress an audio upto 23 TIMES or even more (I haven't tried it myself but you obviously can)!

Your dataset to train the AutoEncoder Model needs to be .wav files since the libraries used in the code are not quite friendly with other formats.

Must note that the compression is for Data Transfer and Data Transmission not Storage since you will get a pickle file which you can decode on the other end! (Both available in the code)

Code of both Encoding and Decoding is available in the jupyter file. Happy Compressing!

If you have a problem getting the dataset, feel free to reach out to me @ waheed@pnw.edu

This project was supervised by Professor Orhan Arikan, Department Chair of Bilkent University's EEE department.