Localization of multiple acoustic sources with small arrays using a coherence test

Area of Science:

• Signal Processing
• Acoustics
• Array Signal Processing

Background:

• Direction finding with fewer sensors than sources is crucial for applications like surveillance and hearing aids.
• Existing methods face challenges with complex acoustic environments and limited sensor configurations.

Purpose of the Study:

• To develop an effective direction-finding algorithm for scenarios with limited sensors.
• To enable the localization of multiple simultaneous time-frequency-sparse sound sources.

Main Methods:

• A coherence test identifies low-rank time-frequency bins for time-frequency-sparse sources.
• Two processing pathways are explored: narrowband spatial spectrum estimation and covariance matrix clustering.
• Experimental validation using omnidirectional and directional microphones.

Main Results:

• Accurate localization of 3-5 simultaneous speech sources with 2-3 microphones (<1 degree error).
• Successful simultaneous localization of two moving military vehicles and gunfire.
• Demonstrated effectiveness over a 4-second duration.

Conclusions:

• The proposed technique offers robust multi-source direction finding with minimal sensor arrays.
• Applicable to diverse acoustic scenarios, including speech and complex environmental sounds.
• Significant advancement for surveillance, teleconferencing, and auditory scene analysis.