UAV sound source distance estimation based on ResNet-CBAM-GRU model
Qing He, Yuhang Chen, Kuangang Fan, Zhiyu Zeng · 2025
In recent years, the rapid development of unmanned aerial vehicle (UAV) technology has enabled its widespread application in various fields. However, with the sharp increase in the number of UAVs, issues such as privacy concerns and airspace security risks caused by their abuse have become increasingly prominent. In the interception of illegal UAV, distance estimation is also a key aspect. In this paper, a ResNet-CBAM-GRU model is proposed for estimating the distance of a UAV sound source. The model uses single-channel audio data as input, employs a residual neural network (ResNet) as the main architecture, combines the convolutional block attention module (CBAM) to focus on key acoustic features, and utilizes a gated recurrent unit (GRU) to capture temporal dynamics, thereby improving distance estimation accuracy. The experimental results show that the mean absolute error (MAE) of this model in the distance estimation of unmanned aerial vehicle sound sources in outdoor environments is 0.5275, and the mean relative error (MRE) is 0.0455. These results verify the effectiveness of the model in estimating the distance of unmanned aerial vehicle sound sources.