Multiple Frequency and Source Angle Estimation by Gaussian Mixture Model with Modified Microphone Array Data Model

Abstract

This paper presents an alternative and efficient one-dimensional direction-of-arrival estimation method for wide-band sources. The proposed method is compatible with most classical subspace-based methods, such as, conventional and root multiple signal classification. Although it only employs a Gaussian mixture model with a maximum likelihood estimation algorithm, it is sufficient for exhibiting wide-band sources angle estimation. Modification of the microphone array data model is proposed and investigated in order to avoid confusion caused by unwanted side lobes in uniform linear arrays radiation. The performance is evaluated in terms of the root-mean-squared error over a range of the signal-to-noise ratio. In conclusion, the proposed method enables the synthesis of signal sources and provides a potential alternative to intelligent source localization systems.