Drone sound detection system based on feature result-level fusion using deep learning
Drone sound detection system based on feature result-level fusion using deep learning
复制标题
基于深度学习特征结果级融合的无人机声音检测系统
DOI：
10.1007/s11042-022-12964-3
复制
发表时间：
2022-06
影响因子：
3.6
通讯作者：
Xiaolin Liu
中科院分区：
文献类型：
--
作者：
Qiushi Dong;Yu Liu;Xiaolin Liu
无人机以其方便快捷、应用广泛等特点受到越来越多的关注，在民用和军用领域发挥着重要的作用。但这些也对人类生命构成威胁，并侵犯隐私，需要在重要的无人值守区域对无人机进行有效且低成本的检测。在本文中，我们提出了结果级融合卷积神经网络（CNN）网络来检测无人机并区分周围环境中是否有无人机。利用对数梅尔谱图和梅尔倒谱系数（MFCC）对声音信号进行特征提取，并将两种特征分别输入到网络中，然后将两种网络的结果用证据理论进行融合，得到最终的检测结果。实验结果表明，基于深度学习方法的无人机检测准确率高于机器学习方法，结果级融合可以联合收割机结合不同特征的优点，准确率提高到94.5%。此外，实验结果表明，所提出的无人机声音检测系统能够在50 m范围内实现有效检测。
Drones have attracted more and more attention due to the convenience and wide applications, and they are playing important roles for both civilian and military. But these also pose threats to human life and invasion to privacy, requiring effective and low-cost detection of drones in important unattended areas. In this paper, we propose the result-level fusion convolutional neural network (CNN) network to detect drones and distinguish whether there are drones in the surrounding environment. Log-Mel spectrogram and Mel frequency cepstral coefficient (MFCC) were used to extract the features of sound signals, and input the two features into the networks separately, then fuse the results from the two networks with evidence theory to obtain the final detection result. The experimental results show that the accuracy of the drone detection based on deep learning method is higher than the machine learning method, the result-level fusion can combine the advantages of different features and increase the accuracy to 94.5%. Furthermore, the results show that the proposed drone sound detection system can achieve effective detection within 50 m.
登录
查看更多内容
影响因子：
7.3
作者：
Xianjun Xia;R. Togneri;Ferdous Sohel;Yuanjun Zhao;David Huang
通讯作者：
Xianjun Xia;R. Togneri;Ferdous Sohel;Yuanjun Zhao;David Huang
DOI：
10.1016/j.comcom.2020.02.065
发表时间：
2020-02
期刊：
ArXiv
影响因子：
--
作者：
Z. Uddin;M. Altaf;M. Bilal;L. Nkenyereye;A. Bashir
通讯作者：
Z. Uddin;M. Altaf;M. Bilal;L. Nkenyereye;A. Bashir
DOI：
10.1109/tmtt.2019.2961911
发表时间：
2020-01
影响因子：
4.3
作者：
Junhyeong Park;Dae-Hwan Jung;Kyung-Bin Bae;Seong‐Ook Park
通讯作者：
Junhyeong Park;Dae-Hwan Jung;Kyung-Bin Bae;Seong‐Ook Park
影响因子：
7.4
作者：
Kim, Jinwoo;Min, Kyungjun;Chi, Seokho
通讯作者：
Chi, Seokho
影响因子：
6.8
作者：
Anwar, Muhammad Zohaib;Kaleem, Zeeshan;Jamalipour, Abbas
通讯作者：
Jamalipour, Abbas