Track and Trace Multiple Incoherent Sound Sources in 3D Space in Real Time

Abstract

This paper presents a methodology for tracking and tracing multiple incoherent sound sources in 3D space in real time. A salient feature of this methodology is its capability of handling all types of sound signals, including broadband, narrowband, continuous, impulsive, and tonal (sinusoidal) sounds over the audible frequency range (20 to 20,000 Hz). Locations of sound sources are indicated in terms of the Cartesian coordinates in real time. The target sources are viewed through an automatic tracking camera covering 350 degree solid angle. The hardware includes four microphones, a thermometer, a webcam, a five-channel signal conditioner and a laptop. Thus, the system can be made light, portable, easy to setup and use and inexpensive. The underlying algorithm is a hybrid approach consisting of modeling of sound radiation from a point source in a free field, triangulation and signal processing techniques. To acquire better understanding of the performance of the device, numerical simulations are conducted to study the impacts of signal noise ratio, microphone spacing, source distance and frequency on the spatial resolution and accuracy of the results. Experiments are carried out to validate results over a wide variety of real-world sound signals such as helicopter noise, human conversations, truck pass-by noise, gun shots, impact sounds, clapping, coughing, etc. Satisfactory results are obtained in most cases, even when a source is behind the measurement microphones.