基于改进MFCC的无人机监测方法

Drone monitoring based on improved MFCC

针对基于光学与无线电的无人机反制设备普遍存在的工艺复杂、制造成本高昂、便携性低等问题。对基于声学的无人机反制方案进行研究。根据无人机螺旋桨的声信号对无人机进行监测,使用梅尔倒谱(MFCC)技术与隐马尔科夫模型(HMM)进行特征提取与分类识别,并针对无人机场景中难以确定和消除的环境噪声,引入掩蔽滤波理论,抑制噪声分量,突出无人机声信号。通过仿真分析验证了基于掩蔽滤波技术改进的MFCC方法具有更高的识别率,并且在低信噪比场景中具有更高的抗噪能力。

In order to solve those problems such as complex process,high manufacturing cost and low portability for Anti-UAV（ Unmanned Aerial Vehicle） system which based on optical and radio. The Anti scheme based on acoustic signal is studied. Monitoring UAV according to the acoustic signal from the propeller. The MFCC（ Mel-frequency cepstral coefficients） technique is used for feature extraction and HMM（ Hidden Markov Mod） for classification. In view of the environmental noise that is difficult to determine and eliminate in the UAV scene,the masking filter theory is introduced to suppress the noise component and to highlight the UAV sound signal. Simulation analysis shows that the improved MFCC method based on masking filter technology has higher recognition rate and has higher noise immunity in low SNR scenarios.