Single-sensor multispeaker listening with acoustic metamaterials

Area of Science:

• Acoustics and Materials Science
• Signal Processing
• Biomedical Engineering

Background:

• The challenge of isolating specific sounds in noisy environments, akin to the human 'cocktail party effect', has long been a goal in auditory research.
• Existing methods often rely on complex algorithms or multiple sensors, limiting practical applications in compact devices.

Purpose of the Study:

• To design and demonstrate a single-sensor device capable of separating simultaneous, overlapping audio sources.
• To achieve this separation using physical layer encoding with acoustic metamaterials, bypassing the need for source-specific characteristics.

Main Methods:

• Utilized a compact array of resonant acoustic metamaterials to create a novel single-sensor listening device.
• Employed principles of compressive sensing and physical layer encoding for audio signal segregation.
• Tested the device's ability to distinguish three independent, overlapping sound sources.

Main Results:

• The developed device successfully separated simultaneous overlapping sounds from three distinct sources.
• Achieved a high correct audio recognition rate of 96.67%.
• Demonstrated that audio segregation was independent of source characteristics, relying solely on physical encoding.

Conclusions:

• This hardware-based approach offers a robust solution for multichannel source separation.
• Potential applications include advanced hearing aids, improved speech recognition systems, and other acoustic sensing technologies.
• The findings pave the way for more sophisticated acoustic imaging and sensing devices.