Three-dimensional reflector localisation and room geometry estimation using a spherical microphone array

Area of Science:

• Acoustics
• Signal Processing
• Computational Geometry

Background:

• Room impulse response analysis is crucial for applications like source localization and acoustic simulation.
• Estimating room geometry from acoustic data, known as geometry inference, is challenging, especially for non-convex spaces.
• Existing methods struggle with the complexity introduced by non-convex room shapes.

Purpose of the Study:

• To present a new geometry inference method capable of localizing reflective boundaries.
• To estimate room geometry for both convex and non-convex environments.
• To evaluate the method's performance using simulated and real-world acoustic data.

Main Methods:

• The proposed method analyzes room impulse responses to identify reflections.
• It utilizes information on the time- and direction-of-arrival of these reflections.
• A spherical microphone array with multiple channels was employed for data acquisition.

Main Results:

• The general shape of tested rooms was successfully inferred using the developed method.
• Higher accuracy was observed in inferring the geometry of convex rooms compared to non-convex ones.
• Inaccuracies were linked to room complexity and errors in estimating reflection arrival times and directions.

Conclusions:

• The presented geometry inference method offers a viable approach for determining room shapes from acoustic data.
• The method demonstrates applicability to both simple and complex room geometries.
• Further improvements in reflection parameter estimation are needed for enhanced accuracy, particularly in complex acoustic environments.