3D Sound Source Mapping Using Interval-Based Methods

Abstract

AbstractThe knowledge about the 3D position of sound sources can expand the model of the environment used by a mobile robot. To describe the 3D position of multiple sound sources, we form a relationship between the extracted auditory features from the microphone signals and the microphones’ position. To link all observations from different measurement positions, we form a Constraint Satisfaction Problem (CSP). In this work interval analysis is used to solve these equations. Compared to classical approaches, which calculate an approximation of this non-linear system of equations and require a good initial guess, our approach consumes less computation effort. As a benefit, our approach enables the calculation at run time on the hardware of a robot. To increase our method’s robustness for real applications, we extended the approach to model the uncertainties of the auditory features and the microphone positions using interval-based methods. For the extraction of the auditory features, we developed two methods. The first method is based on the direction estimations of state-of-the-art approaches. In contrast, the second method - Interval-Based Time Difference (InTiD) - is directly using interval-based methods on the microphones’ acoustics signals. Finally, by using simulated and real data, we demonstrate the functionality of our Interval-Based Sound Source Mapping (IB-SSM).KeywordsMobile roboticsSound source mappingRobot auditionInterval analysis