NMF Techniques for single source audio separation in the BeeMon project

BeeMon [1] is a project developed by students and staff of the Appalachian State University’s Computer Science Department, with the objective of gathering information about the state of monitored beehives with the ultimate goal of helping beekeepers recover the bee populations. For this purpose, audio processing is performed to detect anomalies in the beehives, and so it is of strong interest to be able to automatically separate the sound generated by the beehive itself from other sources of environmental noise, such as human, animal, or meteorological noise. With this work it is intended to contribute to the development of the project through tools and information of value to the aforementioned task. In this document 4 interesting methods will be discussed.