Loading [a11y]/accessibility-menu.js
Improving Source Tracking Accuracy Through Learning-Based Estimation Methods in SH Domain: A Comparative Study | IEEE Journals & Magazine | IEEE Xplore

Improving Source Tracking Accuracy Through Learning-Based Estimation Methods in SH Domain: A Comparative Study


Impact Statement:This research presents a robust framework for acoustic source tracking, employing learning-based DOA estimation methods in the SHs domain. Our comprehensive evaluation ac...Show More

Abstract:

Acoustic source tracking is significant across applications like surveillance, teleconferencing, and robot audition, yet the complexity introduced by reverberation, backg...Show More
Impact Statement:
This research presents a robust framework for acoustic source tracking, employing learning-based DOA estimation methods in the SHs domain. Our comprehensive evaluation across diverse scenarios and filters shows that Kalman filtering achieves over 50% higher tracking accuracy. Notably, our innovative SH learning-based models, such as Spherical Harmonics-INTensity (SH-INT), SH-CRNN, and spherical harmonics-convolutional neural network with matching pursuit (SH-CNN-MP), consistently outperform the baseline multiple signal classification in the SH domain (SH-MUSIC) method, demonstrating their efficacy in source tracking with the error less than 1°. This work significantly contributes to the advancement of intelligent acoustic source tracking, with potential implications for various applications, including surveillance, teleconferencing, and robotics, with potential impacts in industries reliant on acoustic source tracking.

Abstract:

Acoustic source tracking is significant across applications like surveillance, teleconferencing, and robot audition, yet the complexity introduced by reverberation, background noise, and overlapping sources impedes precise source localization. This article uses learning-based localization methods to introduce a resilient and intelligent acoustic source tracking approach in the spherical harmonics (SHs) domain. The tracking algorithms anticipate moving source locations by leveraging past predictions and direction of arrival (DOA) estimations. The prediction probability is computed through alpha–beta and Kalman filtering applied to the estimated DOAs, which are likelihood probabilities obtained from learning models. Utilizing the spatial attributes of sound sources encoded in SH signals, diverse learning-based frameworks are introduced to capture the intricate relationship between SH features and source locations. Supervised learning is utilized to train the models that minimize localiza...
Published in: IEEE Transactions on Artificial Intelligence ( Volume: 5, Issue: 8, August 2024)
Page(s): 3974 - 3984
Date of Publication: 17 January 2024
Electronic ISSN: 2691-4581

Funding Agency:


Contact IEEE to Subscribe

References

References is not available for this document.