[Paper Review] Symphony: Localizing Multiple Acoustic Sources with a Single Microphone Array

Weiguo Wang, Jinming Li+2 more|arXiv (Cornell University)|Sep 30, 2022

Speech and Audio Processing4 citations

TL;DR

Symphony proposes the first method for localizing multiple acoustic sources using a single microphone array, leveraging geometry-based filtering to separate signals by path and coherence-based clustering to group signals from the same source. It achieves a median localization error of 0.694m—68% lower than the state of the art—enabling accurate, real-time source localization on consumer devices without distributed arrays.

ABSTRACT

Sound recognition is an important and popular function of smart devices. The location of sound is basic information associated with the acoustic source. Apart from sound recognition, whether the acoustic sources can be localized largely affects the capability and quality of the smart device's interactive functions. In this work, we study the problem of concurrently localizing multiple acoustic sources with a smart device (e.g., a smart speaker like Amazon Alexa). The existing approaches either can only localize a single source, or require deploying a distributed network of microphone arrays to function. Our proposal called Symphony is the first approach to tackle the above problem with a single microphone array. The insight behind Symphony is that the geometric layout of microphones on the array determines the unique relationship among signals from the same source along the same arriving path, while the source's location determines the DoAs (direction-of-arrival) of signals along different arriving paths. Symphony therefore includes a geometry-based filtering module to distinguish signals from different sources along different paths and a coherence-based module to identify signals from the same source. We implement Symphony with different types of commercial off-the-shelf microphone arrays and evaluate its performance under different settings. The results show that Symphony has a median localization error of 0.694m, which is 68% less than that of the state-of-the-art approach.

Motivation & Objective

• To enable accurate localization of multiple acoustic sources using a single, compact microphone array, overcoming limitations of existing systems.
• To address the challenge of source separation and localization in dense acoustic environments where multiple sources emit simultaneously.
• To eliminate the need for distributed microphone arrays, making the solution viable for consumer smart devices like smart speakers.
• To develop a method that is robust across different microphone array configurations and real-world conditions.
• To achieve high-precision localization with minimal hardware requirements, suitable for commercial off-the-shelf devices.

Proposed method

• Symphony employs a geometry-based filtering module that exploits the fixed spatial configuration of microphones to distinguish signals arriving along different paths from distinct sources.
• It uses the unique time-delay and amplitude relationships across microphones to model signal propagation and separate sources based on their geometric signature.
• A coherence-based module identifies and clusters signals originating from the same source by measuring signal similarity across array elements.
• The method jointly estimates direction-of-arrival (DoA) for each source by combining filtered signals and coherence scores.
• It processes signals in real time using a pipeline that first isolates path-specific components and then groups them by source coherence.
• The approach is implemented on various commercial microphone arrays, demonstrating hardware-agnostic performance across different form factors.

Experimental results

Research questions

• RQ1 Can multiple acoustic sources be accurately localized using only a single microphone array, without requiring a distributed network?
• RQ2 How can signals from different sources be separated when they arrive along overlapping paths in a single array?
• RQ3 What role does the geometric layout of microphones play in distinguishing source-specific signal components?
• RQ4 How does coherence-based grouping improve source localization accuracy in multi-source scenarios?
• RQ5 Can the proposed method achieve superior localization performance compared to state-of-the-art approaches on real-world hardware?

Key findings

• Symphony achieves a median localization error of 0.694 meters across various test conditions, significantly outperforming the state-of-the-art method.
• The median error reduction of 68% compared to the prior SOTA demonstrates substantial improvement in localization accuracy.
• The method maintains consistent performance across different types of commercial off-the-shelf microphone arrays, indicating robustness to hardware variation.
• The geometry-based filtering effectively separates signals from different sources even when they arrive from similar directions.
• The coherence-based clustering successfully groups signals from the same source, enabling accurate DoA estimation per source.
• The system operates in real time and is deployable on standard smart devices such as Amazon Alexa, enabling practical integration.