Application of deep learning for accurate source localization using sound intensity vector

Iljoo Jeong; In-Jee Jung; Seungchul Lee

doi:10.7776/ASK.2024.43.1.072

All Issue

2024 Vol.43, Issue 1 Preview Page Next Page

Research Article

Application of deep learning for accurate source localization using sound intensity vector 음향인텐시티 벡터를 통해 정확한 음원 위치 추정을 위한 딥러닝 적용

31 January 2024. pp. 72-77

PDF XML

Abstract

Recently, the necessity for sound source localization has grown significantly across various industrial sectors. Among the sound source localization methods, sound intensimetry has the advantage of having high accuracy even with a small microphone array. However, the increase in localization error at high Helmholtz numbers have been pointed out as a limitation of this method. The study proposes a method to compensate for the bias error of the measured sound intensity vector according to the Helmholtz numbers by applying deep learning. The method makes it possible to estimate the accurate direction of arrival of the source by applying a dense layer-based deep learning model that derives compensated sound intensity vectors when inputting the sound intensity vectors measured by a tetrahedral microphone array for the Helmholtz numbers. The model is verified based on simulation data for all sound source directions with 0.1 < kd < 3.0. One can find that the deep learning-based approach expands the measurement frequency range when implementing the sound intensimetry-based sound source localization method, also one can make it applicable to various microphone array sizes.

Keywords

Sound source localization

Deep-learning

Sound intensimetry

Sound intensity vector compensation

최근 여러 산업 분야에서 음원 위치 추정의 필요성이 커지고 있다. 기존 음원 위치 추정 방법들 중에서, 음향 인텐시티 계측법은 작은 마이크로폰 어레이에서도 높은 정확도를 가지는 장점이 있다. 그러나, 높은 헬름홀츠 수에서의 위치 추정 오차 증가는 이 방법의 한계로 지적되어 왔다. 본 연구에서는 이러한 헬름홀츠 수에 따른 인텐시티 편향 오차를 딥러닝을 통해 보상하는 방법을 제안한다. 제안된 방법은 정사면체 마이크로폰 어레이에서 헬름홀츠 수에 대해 측정된 음향인텐시티 벡터를 입력했을 때, 보상된 음향 인텐시티 벡터를 도출하는 밀집 층 기반의 딥러닝 모델을 적용하여 정확한 음원 위치의 추정을 가능케 한다. 본 연구의 제안 모델은, 0.1 < kd < 3.0의 모든 음원 방향에 대한 시뮬레이션 데이터를 기반으로 검증하였다. 이를 통해, 딥러닝 기반 접근 방식은 음향 인텐시티 벡터 기반의 음원 추정법을 적용하는데 있어서 측정 주파수 범위를 확장하고 다양한 크기를 갖는 마이크로폰 어레이에 적용할 수 있음을 확인하였다.

키워드

음원 위치 추정

딥러닝

음향인텐시티법

음향 인텐시티 벡터 보상

References

P. Chiariotti, M. Martarelli, and P. Castellini, "Acoustic beamforming for noise source localization-reviews, methodology and applications," MSSP, 120, 422-448 (2019). 10.1016/j.ymssp.2018.09.019

C. Sun and Y. Liu, "Spherical reverse beamforming for sound source localization based on the inverse method," Sensors, 19, 2618 (2019). 10.3390/s1911261831181841PMC6603625

J. H. DiBiase, H. F. Silverman, and M. S. Brandstein, Microphone Arrays (Springer Berlin, Heidelberg, 2001), pp. 157-180. 10.1007/978-3-662-04619-7_8

M. Risoud, J.-N. Hanson, F. Gauvrit, C. Renard, P.-E. Lemesre, N.-X. Bonne, and C. Vincent, "Sound source localization," Eur. Ann. Otorhinolaryngol. Head Neck Dis. 135, 259-264 (2018). 10.1016/j.anorl.2018.04.00929731298

G. Pavic, "Measurement of sound intensity," J. Sound Vib. 51, 533-545 (1977). 10.1016/S0022-460X(77)80050-3

I.-J Jung and J.-G Ih, "Comparison of the sound source localization methods appropriate for a compact microphone array" (in Korean), J. Acoust. Soc. Kr. 31, 47-56 (2020).

I.-J Jung and J.-G Ih, "Compensation of inherent bias errors in using the three-dimensional acoustic intensimetry for sound source localization," J. Sound Vib. 461, 114918 (2019). 10.1016/j.jsv.2019.114918

I.-J Jung and J.-G Ih,, "Combined microphone array for precise localization of sound source using the acoustic intensimetry," MSSP. 160, 107820 (2021). 10.1016/j.ymssp.2021.107820

A. Kujawski, G. Herold, and E. Sarradj, "A deep learning method for grid-free localization and quantification of sound sources," J. Acoust. Soc. Am. 146, EL225-EL231 (2019). 10.1121/1.512602031590523

S. Y. Lee, J. Chang, and S. Lee, "Deep learning-based method for multiple sound source localization with high resolution and accuracy," MSSP. 161, 107959 (2021). 10.1016/j.ymssp.2021.107959

J.-C. Pascal and J.-F. Li, "A systematic method to obtain 3D finite-difference formulations for acoustic intensity and other energy quantities," J. Sound Vib. 310, 1093-1111 (2008). 10.1016/j.jsv.2007.08.029

D. P. Kingma and J. L. Ba, "Adam: A method for stochastic optimization," arXiv preprint arXiv:1412. 6980 (2014).

Information

Publisher :The Acoustical Society of Korea
Publisher(Ko) :한국음향학회
Journal Title :The Journal of the Acoustical Society of Korea
Journal Title(Ko) :한국음향학회지
Volume : 43
No :1
Pages :72-77
Received Date : 2023-12-04
Accepted Date : 2023-12-29
DOI :https://doi.org/10.7776/ASK.2024.43.1.072

The Journal of the Acoustical Society of KoreaISSN:1225-4428(Print) 2287-3775(Online)한국음향학회

All Issue