A New Convolutional Mapping DNN for Sound Source Localization Using Microphone Arrays

Sound source localization in noisy environments using microphone arrays is a challenging task that has attracted the attention of many researchers. Using deep learning approaches by reformulating the sound source localization (SSL) obtained superior performances in these kinds of classification tasks. In this work, a new model of a convolutional neural network with specific extracted features is introduced to infer the direction of arrival (DOA) of a sound source in noisy conditions. Three types of experiments (i.e., the volume of room, the distance of the sources, and the position of the microphone) are designed to examine the generalizability of the proposed model. The results depicted that the proposed model is able to perform on the first two conditions by achieving the highest accuracy in DOA estimation.

Keywords:

Convolutional Neural Network , Deep Learning , Microphone Arrays , Feature Extraction , Sound Source Localization.