Detection and recognition of multiple types of small rotary-wing drones based on optical fiber EFPI acoustic sensor
提出一种基于石墨烯膜片的非本征型光纤法布里-珀罗干涉仪(extrinsic Fabry-Perot interferometer based on optical fiber,EFPI)声传感器,结合卷积神经网络(convolutional neural networks, CNN)和长短期记忆网络(long short term memory network, LSTM)级联模型,用于多种类小型旋翼无人机探测与识别,并对实验过程中的提取特征、采集装置、声源种类数量进行了进一步的细化分析。通过实验对比,发现用梅尔倒谱系数(Mel frequency cepstral coefficient, MFCC)作为特征提取时,识别效果明显优于短时傅里叶变换(short-time Fourier transform,STFT)、梅尔频谱(Mel spectrogram)特征,而且使用该传感器进行无人机探测与识别的效果比电学式麦克风的准确率高约2%。此外,对多种类小型旋翼无人机声源的识别准确率均在95.33%以上,证实该方法可对多类无人机进行有效探测识别。
Authors
- Lin Mei Yang (ORCID: https://orcid.org/0000-0002-4148-5740)
- Bin Liu (ORCID: https://orcid.org/0000-0002-1566-6302)
- DING Rou
- ZHAO Qichao
- WANG Haoqi
- REN Hongzhao
- XIAO Shuang
Publication Details
- Journal
- Journal of Applied Optics
- Published
- 2025-01-01
- DOI
- https://doi.org/10.5768/jao202546.0311003
- Primary Topic
- Advanced Sensor and Control Systems
- Type
- article
- Field-Weighted Citation Impact
- 0.00