A new complex mapping method of neural networks used in sound source localization
Source abstract
Abstract
Sound source localization has a wide range of application prospects in many fields, such as smart home and audio monitoring. Traditional methods are difficult to achieve accurate location in the face of multi-path reflection, reverberation, and ambient noise. In this paper, a complex mapping conversion method for sound source location is proposed. By using complex-valued convolutional neural networks to fuse the amplitude and phase information of the data, a more accurate and comprehensive analysis can be carried out to improve its robustness and realize the accurate location of the sound source. The sound source location method based on complex-valued convolutional neural networks is studied, and the complex mapping principle is analyzed. Simulation and experimental studies were carried out, and the results of simulation and experiment are basically consistent. In the experiment, the positioning accuracy of the complex mapping method is 9.49% higher than that of the absolute value method and 15.81% higher than that of the phase angle method. In addition, its localization success rate, respectively, increased by 4.9% and 8.6% compared to two other methods. This paper opens up a new way for the application of complex-valued convolutional neural networks in sound source localization.
Source: crossref · Quality: baseline
Research context
OpenAlex topics: Speech and Audio Processing · Underwater Acoustics Research · Music and Audio Processing
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
Affiliations
Hubei Key Laboratory of Modern Manufacturing Quantity Engineering, School of Mechanical Engineering, Hubei University of Technology , Wuhan, Hubei 430068, China
Funding
National Natural Science Foundation of China · Hubei Provincial Natural Science Foundation of China · Science and Technology Research Project of Education Department of Hubei Province · Green Industry Technology Leading Project of Hubei University of Technology
References (27)
Show source-provided references
1. A survey of sound source localization with deep learning methods10.1121/10.0011809 ↗
2. A review on sound source localization systems10.1007/s11831-022-09747-2 ↗
3. Acoustic emission source location method and experimental verification for structures containing unknown empty areas10.1016/j.ijmst.2022.01.002 ↗
4. A sound source localization method based on improved second correlation time delay estimation10.1088/1361-6501/aca5a6 ↗
5. An end-to-end sound source navigation method of indoor mobile robot10.1109/CCDC49329.2020.9164238 ↗
6. RoboCup@Home: Evaluation of voice recognition systems for domestic service robots and introducing Latino Dataset10.1109/LARS/SBR/WRE54079.2021.9605485 ↗
7. Sound source localization–state of the art and new inverse scheme10.1007/s00502-021-00881-6 ↗
8. A sound source localization method based on frequency divider and time difference of arrival10.3390/app13106183 ↗
9. Multiple sound sources localization from coarse to fine10.1007/978-3-030-58565-5_18 ↗
10. Comparison of convolution types in CNN-based feature extraction for sound source localization10.23919/Eusipco47968.2020.9287344 ↗
11. A neural network based microphone array approach to grid-less noise source localization10.1016/j.apacoust.2021.107947 ↗
12. On improved training of CNN for acoustic source localisation10.1109/TASLP.2021.3049337 ↗
13. On the appropriateness of complex-valued neural networks for speech enhancement10.21437/Interspeech.2016-300 ↗
14. C. Trabelsi , O.Bilaniuk, Y.Zhang, “Deep complex networks,” arXiv:170509792 (2017).
15. Complex spectrogram enhancement by convolutional neural network with multi-metrics learning10.1109/MLSP.2017.8168119 ↗
16. An approach for sound source localization by complex-valued neural network10.1587/transinf.e96.d.2257 ↗
17. Sound source localization using a convolutional neural network and regression model10.3390/s21238031 ↗
18. Deep learning based multi-source localization with source splitting and its effectiveness in multi-talker speech recognition10.1016/j.csl.2022.101360 ↗
19. J. Bassey , L.Qian, and X.Li, “A survey of complex-valued neural networks,” arXiv:210112249 (2021).
20. Olver · 2017 · 806
21. Acoustic source localization in the circular harmonic domain using deep learning architecture10.1109/taslp.2022.3190723 ↗
22. Chakrabarty · Robust drection-of-arrival estimation and spatial filtering in noisy and reverberant environments · 2020
23. Acoustic localization enhanced with phase information from modified STFT magnitude10.1109/ICDSP.2013.6622673 ↗
24. S-DCCRN: Super wide band DCCRN with learnable complex feature for speech enhancement10.1109/ICASSP43922.2022.9747029 ↗
25. Complex-valued neural networks: A comprehensive survey10.1109/jas.2022.105743 ↗
26. “Real-valued backpropagation is unsuitable for complex-valued neural networks,” in Proceedings of Advances in Neural Information Processing Systems 35 - 36th Conference on Neural Information Processing Systems
27. Understanding of a convolutional neural network10.1109/ICEngTechnol.2017.8308186 ↗