Sound source localization: microphone array design and evolutionary estimation

The use of speech in human-robot interaction is an indispensable feature from the points of view of both the human and the robot. In order to realize such capability, sound source localization becomes a mandatory or a pre-requisite requirement. This paper describes the localization of a speaker by using a microphone array. Time delay of arrival (TDOA) based technique is employed to obtain the angular direction of the sound source relative to the array. Geometric array construction that facilitates sound source localization is taken into consideration. Since sound reception is corrupted by noise, an evolutionary computation method (genetic algorithms) is also used to provide an improved estimation of the sound source location. A simplification of the genetic algorithms implementation is carried out, such that the complexity of the calculation is reduced and makes the estimator practically tractable. By using the proposed method, the need for an accurate measurement model and a speaker motion model is removed and thus the estimation robustness is increased. Simulations are included to demonstrate the effectiveness of the proposed technique.