Summer Practice at Aselsan Inc.

2019-09-03

Supervised by Lütfi Murat Gevrekçi, PhD

---

TODO: from final report

In this summer practice, I was given the task to explore two new scientific papers \cite{doanet}\cite{seldnet}, that focus on deep neural network (DNN) applications of direction-of-arrival (DOA) estimation and sound event detection (SED) task. I was asked to explore the success and integrability of these approaches to ongoing projects. Both published in 2018, the papers mainly feature the works of Audio and Signal Processing Research Group in Tampere University, Finland. To fully understand the concepts and approaches presented in the papers, I studied the basics of machine learning and neural network concepts. I reviewed popular approaches to DOA estimation and SED task, being DNN-based or not. I also set the computer environment up, to train and test the neural network suggested in papers with the dataset provided, and observe the results. After my research, I gave a presentation to my superiors about these papers, and newer approaches to DOA estimation and SED task. I introduced them the DCASE community, abbreviation for \textit{Detection, Classification of Acoustic Scenes and Events}, which is a community that organizes challenges annually on aspects of environmental sound classification and detection.