Deep Learning for Audio Signal Source Positioning Using Microphone Array

摘要

This paper deals with a deep learning of audio signal source positioning. When there is no cooperation between signal source and receiving microphones, it is best way to employ the time difference of arrival (TDOA) task for source positioning. Therefore, this paper considers positioning using a microphone array and TDOA method. Alongside with analytical methods, such as triangulation, this study concentrates on statistical signal processing and deep learning by neural network. Considering the location finding applications the idea of using a neural network algorithm is quite new and novel, but its performance was questionable before presented study. This study shows that neural network type deep learning algorithm performs better over analytical techniques and provides much faster location estimate, as it is presented in this study.