Direction of Arrival Estimation for Multiple Sound Sources Using Convolutional Recurrent Neural Network

Sharath Adavanne; Archontis Politis; Tuomas Virtanen

doi:10.23919/EUSIPCO.2018.8553182

Direction of Arrival Estimation for Multiple Sound Sources Using Convolutional Recurrent Neural Network

Sharath Adavanne^*, Archontis Politis, Tuomas Virtanen

^*Corresponding author for this work

Tampere University of Technology

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

185 Citations (Scopus)

Abstract

This paper proposes a deep neural network for estimating the directions of arrival (DOA) of multiple sound sources. The proposed stacked convolutional and recurrent neural network (DOAnet) generates a spatial pseudo-spectrum (SPS) along with the DOA estimates in both azimuth and elevation. We avoid any explicit feature extraction step by using the magnitudes and phases of the spectrograms of all the channels as input to the network. The proposed DOAnet is evaluated by estimating the DOAs of multiple concurrently present sources in anechoic, matched and unmatched reverberant conditions. The results show that the proposed DOAnet is capable of estimating the number of sources and their respective DOAs with good precision and generate SPS with high signal-to-noise ratio.

Original language	English
Title of host publication	2018 26TH EUROPEAN SIGNAL PROCESSING CONFERENCE (EUSIPCO)
Publisher	IEEE
Pages	1462-1466
Number of pages	5
Volume	2018-September
ISBN (Print)	978-90-827970-1-5
DOIs	https://doi.org/10.23919/EUSIPCO.2018.8553182
Publication status	Published - 2018
MoE publication type	A4 Conference publication
Event	European Signal Processing Conference - Rome, Italy Duration: 3 Sept 2018 → 7 Sept 2018 Conference number: 26

Publication series

Name	European Signal Processing Conference
Publisher	IEEE
ISSN (Print)	2076-1465
ISSN (Electronic)	2076-1465

Conference

Conference	European Signal Processing Conference
Abbreviated title	EUSIPCO
Country/Territory	Italy
City	Rome
Period	03/09/2018 → 07/09/2018

Access to Document

10.23919/EUSIPCO.2018.8553182

https://arxiv.org/pdf/1710.10059.pdf

Cite this

@inproceedings{fc15bc6633dc4d14b7ba8fd1e217711e,

title = "Direction of Arrival Estimation for Multiple Sound Sources Using Convolutional Recurrent Neural Network",

abstract = "This paper proposes a deep neural network for estimating the directions of arrival (DOA) of multiple sound sources. The proposed stacked convolutional and recurrent neural network (DOAnet) generates a spatial pseudo-spectrum (SPS) along with the DOA estimates in both azimuth and elevation. We avoid any explicit feature extraction step by using the magnitudes and phases of the spectrograms of all the channels as input to the network. The proposed DOAnet is evaluated by estimating the DOAs of multiple concurrently present sources in anechoic, matched and unmatched reverberant conditions. The results show that the proposed DOAnet is capable of estimating the number of sources and their respective DOAs with good precision and generate SPS with high signal-to-noise ratio.",

author = "Sharath Adavanne and Archontis Politis and Tuomas Virtanen",

year = "2018",

doi = "10.23919/EUSIPCO.2018.8553182",

language = "English",

isbn = "978-90-827970-1-5",

volume = "2018-September",

series = "European Signal Processing Conference",

publisher = "IEEE",

pages = "1462--1466",

booktitle = "2018 26TH EUROPEAN SIGNAL PROCESSING CONFERENCE (EUSIPCO)",

address = "United States",

note = "European Signal Processing Conference, EUSIPCO ; Conference date: 03-09-2018 Through 07-09-2018",

}

Adavanne, S, Politis, A & Virtanen, T 2018, Direction of Arrival Estimation for Multiple Sound Sources Using Convolutional Recurrent Neural Network. in 2018 26TH EUROPEAN SIGNAL PROCESSING CONFERENCE (EUSIPCO). vol. 2018-September, 8553182, European Signal Processing Conference, IEEE, pp. 1462-1466, European Signal Processing Conference, Rome, Italy, 03/09/2018. https://doi.org/10.23919/EUSIPCO.2018.8553182

Direction of Arrival Estimation for Multiple Sound Sources Using Convolutional Recurrent Neural Network. / Adavanne, Sharath; Politis, Archontis; Virtanen, Tuomas.
2018 26TH EUROPEAN SIGNAL PROCESSING CONFERENCE (EUSIPCO). Vol. 2018-September IEEE, 2018. p. 1462-1466 8553182 (European Signal Processing Conference).

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

TY - GEN

T1 - Direction of Arrival Estimation for Multiple Sound Sources Using Convolutional Recurrent Neural Network

AU - Adavanne, Sharath

AU - Politis, Archontis

AU - Virtanen, Tuomas

N1 - Conference code: 26

PY - 2018

Y1 - 2018

N2 - This paper proposes a deep neural network for estimating the directions of arrival (DOA) of multiple sound sources. The proposed stacked convolutional and recurrent neural network (DOAnet) generates a spatial pseudo-spectrum (SPS) along with the DOA estimates in both azimuth and elevation. We avoid any explicit feature extraction step by using the magnitudes and phases of the spectrograms of all the channels as input to the network. The proposed DOAnet is evaluated by estimating the DOAs of multiple concurrently present sources in anechoic, matched and unmatched reverberant conditions. The results show that the proposed DOAnet is capable of estimating the number of sources and their respective DOAs with good precision and generate SPS with high signal-to-noise ratio.

AB - This paper proposes a deep neural network for estimating the directions of arrival (DOA) of multiple sound sources. The proposed stacked convolutional and recurrent neural network (DOAnet) generates a spatial pseudo-spectrum (SPS) along with the DOA estimates in both azimuth and elevation. We avoid any explicit feature extraction step by using the magnitudes and phases of the spectrograms of all the channels as input to the network. The proposed DOAnet is evaluated by estimating the DOAs of multiple concurrently present sources in anechoic, matched and unmatched reverberant conditions. The results show that the proposed DOAnet is capable of estimating the number of sources and their respective DOAs with good precision and generate SPS with high signal-to-noise ratio.

U2 - 10.23919/EUSIPCO.2018.8553182

DO - 10.23919/EUSIPCO.2018.8553182

M3 - Conference article in proceedings

SN - 978-90-827970-1-5

VL - 2018-September

T3 - European Signal Processing Conference

SP - 1462

EP - 1466

BT - 2018 26TH EUROPEAN SIGNAL PROCESSING CONFERENCE (EUSIPCO)

PB - IEEE

T2 - European Signal Processing Conference

Y2 - 3 September 2018 through 7 September 2018

ER -

Direction of Arrival Estimation for Multiple Sound Sources Using Convolutional Recurrent Neural Network

Abstract

Publication series

Conference

Access to Document

Fingerprint

Cite this