Explore Audio Compression Using Machine Learning

Abstract

Storage requirements for digital applications are increasing rapidly. There are considerable amount of applications in the world that use audio data. Data compression is one of the key solutions in saving storage space, and has been used for many decades. In this paper we explore the possibility of using machine learning dimensionality reduction methods to compress and represent an original audio data file in storage. Reduced audio file will then be compared with the original audio file. In this research we only focus on speech audio data.