Bag-of-Features Methods for Acoustic Event Detection and Classification

Years ago, my colleague Rene Grzeszick and I began collaborating on Acoustic Event Detection (AED) using the Bag-of-Features (BoF) method. It all started because I wanted to exclude footsteps and sounds of moving devices on a table in my speaker localization a geometry calibration methods. After our first joint paper became one of our most cited, we continued to refine and apply the method.
Among other things, I managed to apply it for speech enhancement with Sharon Gannot during my stay at the Department of Electrical Engineering, Bar Ilan University, Ramat Gan, Israel.
Later, I was happy to present our extension for Acoustic Sensor Networks (ASNs) at the fabulous DCASE 2016 Workshop in Budapest, Hungary.
I am happy to announce that the current state of the development along with some extensive evaluation can now be found in a journal paper. We even made it on the cover 😉

Bag-of-Features Methods for Acoustic Event Detection and Classification
Rene Grzeszick, Axel Plinge,Gernot A. Fink
IEEE/ACM Transactions on Audio, Speech, and Language Processing 25(6), pages 1242-1252, June 2017

If you would like to try it out, you can download the Python code on github.

If you are interested in the topic, make sure to check out the upcoming DCASE 2017 Workshop in Munich, Germay.

As usual, there is an accompanying video: