Ensemble NASNet Deep Feature Generator Based Underwater Acoustic Classification Model

Abstract

Sualtı akustiği, gelişmiş sinyal işleme için önemli ve karmaşık araştırma alanlarından biridir. Bu çalışmada, su altı araçlarının yön tespiti için derin öğrenme ve makine öğrenmesi tabanlı bir yöntem önerilmiştir. Sualtı robotu kullanılarak yeni bir veri seti toplandı. Veri setini oluşturmak için su altına bir mikrofon yerleştirildi. Su altında hareket eden Uzaktan Kumandalı Sualtı Aracının (ROV) pervanelerinin sesleri toplandı. Önce su altında hareketsiz bir ses kaydı yapıldı. Sualtı robotu x, y ve z eksenleri boyunca hareket etti ve bir ses veri seti oluşturuldu. Bu veri seti toplamda dört sınıftan oluşmaktadır. Bu sesler üzerinde özellik çıkarımı için NASNetLarge ve NASNetMobile derin öğrenme modelleri uygulanmıştır. Bu iki derin öğrenme modelinin özellikleri birleştirilmiştir. Elde edilen öznitelikler arasından en ağırlıklı öznitelikleri seçmek için chi2 yöntemi kullanılmıştır. Ardından, seçilen özellikleri sınıflandırmak için Destek Vektör Makinesi (SVM) algoritması kullanılır. Sınıflandırmada Doğrusal SVM algoritması ile %77,66 doğruluk hesaplanmıştır.

Keywords

• Sualtı Ses Sınıflandırması
• Sualtı Yön Algılama
• SVM
• NASNetLarge
• NASNetMobile.

Ensemble NASNet Deep Feature Generator Based Underwater Acoustic Classification Model

Abstract

Underwater acoustics is one of the important and complex research areas for advanced signal processing. In this study, a deep learning and machine learning-based method are proposed for direction detection of underwater vehicles. A new dataset was collected using the underwater robot. A microphone was placed underwater to create the dataset. The sounds of the propellers of the Remotely Operated Underwater Vehicle (ROV) moving underwater were collected. First, a still sound recording was made underwater. The underwater robot moved along the x, y, and z axes, and a sound dataset was created. This data set consists of four classes in total. NASNetLarge and NASNetMobile deep learning models have been applied for feature extraction on these sounds. The features of these two deep learning models are combined. The chi2 method was used to select the most weighted features among the obtained features. Then, the Support Vector Machine (SVM) algorithm is used to classify the selected features. In classification, 77.66% accuracy was calculated with the Linear SVM algorithm.

Keywords

• Underwater Sound Classification
• Underwater Direction Detection
• SVM
• NASNetLarge
• NASNetMobile.

References

1. Choi J, Park J, Lee Y, Jung J, Choi HT. Robust directional angle estimation of underwater acoustic sources using a marine vehicle. Sensors (Switzerland) 2018;18. https://doi.org/10.3390/s18003062.
2. Koseoglu M, Bereketli A, Yazgi I, Yeni B. Probabilistic broadcast for dense AUV networks. Ocean 2016 MTS/IEEE Monterey, OCE 2016 2016. https://doi.org/10.1109/OCEANS.2016.7761118.
3. Cho H, Gu J, Joe H, Asada A, Yu SC. Acoustic beam profile-based rapid underwater object detection for an imaging sonar. J Mar Sci Technol 2015;20:180–97. https://doi.org/10.1007/s00773-014-0294-x.
4. Lee H, Jung HK, Cho SH, Kim Y, Rim H, Lee SK. Real-Time Localization for Underwater Moving Object Using Precalculated DC Electric Field Template. IEEE Trans Geosci Remote Sens 2018;56:5813–23. https://doi.org/10.1109/TGRS.2018.2826556.
5. Nie D, Sun Z, Qiao G, Liu S, Yin Y. Kite-type passive acoustic detection system for underwater small targets. 2014 Ocean - St John’s, Ocean 2014 2015. https://doi.org/10.1109/OCEANS.2014.7003207.
6. Isbitiren G, Akan OB. Three-dimensional underwater target tracking with acoustic sensor networks. IEEE Trans Veh Technol 2011;60:3897–906. https://doi.org/10.1109/TVT.2011.2163538.
7. Jiang J, Wu Z, Lu J, Huang M, Xiao Z. Interpretable features for underwater acoustic target recognition. Meas J Int Meas Confed 2020:108586. https://doi.org/10.1016/j.measurement.2020.108586.
8. Neves G, Ruiz M, Fontinele J, Oliveira L. Rotated object detection with forward-looking sonar in underwater applications. Expert Syst Appl 2020;140:112870. https://doi.org/10.1016/j.eswa.2019.112870.

9. Reis CDG, Padovese LR, de Oliveira MCF. Automatic detection of vessel signatures in audio recordings with spectral amplitude variation signature. Methods Ecol Evol 2019;10:1501–16. https://doi.org/10.1111/2041-210X.13245.
10. Choi J, Choi HT. Multi-target localization of underwater acoustic sources based on probabilistic estimation of direction angle. MTS/IEEE Ocean 2015 - Genova Discov Sustain Ocean Energy a New World 2015. https://doi.org/10.1109/OCEANS-Genova.2015.7271437.
11. Sierra E, Contreras J. Classification of small boats using fuzzy classifier. Annu Conf North Am Fuzzy Inf Process Soc - NAFIPS 2015;2015-Septe:0–4. https://doi.org/10.1109/NAFIPS-WConSC.2015.7284174.
12. Fischell EM, Viquez O, Schmidt H. Passive acoustic tracking for behavior mode classification between surface and underwater vehicles. IEEE Int Conf Intell Robot Syst 2018:2383–8. https://doi.org/10.1109/IROS.2018.8593981.
13. Santos-Domínguez D, Torres-Guijarro S, Cardenal-López A, Pena-Gimenez A. ShipsEar: An underwater vessel noise database. Appl Acoust 2016;113:64–9. https://doi.org/10.1016/j.apacoust.2016.06.008.
14. Sutin A, Bunin B, Sedunov A, Sedunov N, Fillinger L, Tsionskiy M, et al. Stevens passive acoustic system for underwater surveillance. 2010 Int Waterside Secur Conf WSS 2010 2010. https://doi.org/10.1109/WSSC.2010.5730286.
15. [15] CHASING | GLADIUS MINI - 4K Underwater Drone with Camera - Chasing Innovation n.d. https://www.chasing.com/gladius-mini.html (accessed November 27, 2020).
16. Saxen F, Werner P, Handrich S, Othman E, Dinges L, Al-Hamadi A. Face attribute detection with mobilenetv2 and nasnet-mobile. Int Symp Image Signal Process Anal ISPA 2019;2019-Septe:176–80. https://doi.org/10.1109/ISPA.2019.8868585.
17. Zoph B, Vasudevan V, Shlens J, Le Q V. Learning Transferable Architectures for Scalable Image Recognition Barret. Proc IEEE Conf Comput Vis Pattern Recognit 2018:8697–710.
18. Baygin N, Baygin M, Karakose M. A SVM-PSO Classifier for Robot Motion in Environment with Obstacles. 2019 Int. Conf. Artif. Intell. Data Process. Symp. IDAP 2019, 2019. https://doi.org/10.1109/IDAP.2019.8875921.
19. [19] Tuncer T, Ertam F. Neighborhood component analysis and reliefF based survival recognition methods for Hepatocellular carcinoma. Phys A Stat Mech Its Appl 2020;540:123143. https://doi.org/10.1016/j.physa.2019.123143.
20. Tsanas A, Little MA, McSharry PE, Spielman J, Ramig LO. Novel speech signal processing algorithms for high-accuracy classification of Parkinsons disease. IEEE Trans Biomed Eng 2012;59:1264–71. https://doi.org/10.1109/TBME.2012.2183367.
21. Singh M, Shaik AG. Faulty bearing detection, classification and location in a three-phase induction motor based on Stockwell transform and support vector machine. Meas J Int Meas Confed 2019;131:524–33. https://doi.org/10.1016/j.measurement.2018.09.013.
22. Yaman O. An automated faults classification method based on binary pattern and neighborhood component analysis using induction motor. Meas J Int Meas Confed 2021;168:108323. https://doi.org/10.1016/j.measurement.2020.108323.
23. Lahmiri S. Parkinson’s disease detection based on dysphonia measurements. Phys A Stat Mech Its Appl 2017;471:98–105. https://doi.org/10.1016/j.physa.2016.12.009.