TY - GEN
T1 - Differentiable Tracking-Based Training of Deep Learning Sound Source Localizers
AU - Adavanne, Sharath
AU - Politis, Archontis
AU - Virtanen, Tuomo
N1 - jufoid=72074
PY - 2021
Y1 - 2021
N2 - Data-based and learning-based sound source localization (SSL) has shown promising results in challenging conditions, and is commonly set as a classification or a regression problem. Regression-based approaches have certain advantages over classification-based, such as continuous direction-of-arrival estimation of static and moving sources. However, multi-source scenarios require multiple regressors without a clear training strategy up-to-date, that does not rely on auxiliary information such as simultaneous sound classification. We investigate end-to-end training of such methods with a technique recently proposed for video object detectors, adapted to the SSL setting. A differentiable network is constructed that can be plugged to the output of the localizer to solve the optimal assignment between predictions and references, optimizing directly the popular CLEAR-MOT tracking metrics. Results indicate large improvements over directly optimizing mean squared errors, in terms of localization error, detection metrics, and tracking capabilities.
AB - Data-based and learning-based sound source localization (SSL) has shown promising results in challenging conditions, and is commonly set as a classification or a regression problem. Regression-based approaches have certain advantages over classification-based, such as continuous direction-of-arrival estimation of static and moving sources. However, multi-source scenarios require multiple regressors without a clear training strategy up-to-date, that does not rely on auxiliary information such as simultaneous sound classification. We investigate end-to-end training of such methods with a technique recently proposed for video object detectors, adapted to the SSL setting. A differentiable network is constructed that can be plugged to the output of the localizer to solve the optimal assignment between predictions and references, optimizing directly the popular CLEAR-MOT tracking metrics. Results indicate large improvements over directly optimizing mean squared errors, in terms of localization error, detection metrics, and tracking capabilities.
KW - Training
KW - Location awareness
KW - Measurement
KW - Deep learning
KW - Direction-of-arrival estimation
KW - Conferences
KW - Training data
KW - sound source localization
KW - deep-learning acoustic processing
KW - multi-target tracking
U2 - 10.48550/arXiv.2111.00030
DO - 10.48550/arXiv.2111.00030
M3 - Conference contribution
T3 - IEEE Workshop on Applications of Signal Processing to Audio and Acoustics
SP - 211
EP - 215
BT - 2021 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics (WASPAA)
PB - IEEE
T2 - IEEE Workshop on Applications of Signal Processing to Audio and Acoustics
Y2 - 17 October 2021 through 20 October 2021
ER -