Computation of spherical sector harmonics norm with application to blind source separation

Area of Science:

• Acoustics
• Signal Processing
• Machine Learning

Background:

• Spherical microphone arrays (SMAs) are effective for source localization and separation.
• Full SMAs are often uneconomical for environments with restricted source regions.

Purpose of the Study:

• To introduce a spherical sector microphone array (SSMA) for blind source separation in restricted environments.
• To develop a novel mathematical framework for source separation using SSMA.

Main Methods:

• Utilized spherical sector harmonics basis function to compute the norm for mixing matrix estimation.
• Employed the mean-shift algorithm for clustering estimated steering vectors.
• Automated source number estimation based on the number of identified clusters.

Main Results:

• Successfully implemented blind source separation using an SSMA for the first time.
• Demonstrated accurate source number estimation through clustering.
• Validated the mathematical framework via simulations and real-world experiments.

Conclusions:

• The SSMA offers a practical and economical solution for source separation in constrained acoustic spaces.
• The proposed method effectively identifies and separates sound sources with automatic source counting.
• The framework shows robustness and applicability in diverse acoustic scenarios.