Evaluation of Beamforming Techniques for Speaker Identification

This paper analyses DAS, MVDR and MUSIC beamforming algorithms for direction of arrival estimation regarding the usage on mobile robot platforms with the objective to identify a possible communication partner. With this background, the beamforming algorithms are described, analyzed and evaluated in laboratory environment. The evaluation bases on the accuracy of direction of arrival estimation as a function of observation time and FFT resolution. The paper shows differences and advantages of the distinct algorithms and identifies optimized parameters for the corresponding beamformer used for speaker localization on mobile robot platforms with minimized hardware and space requirements.