Cellphone-Based sUAS Range Estimation: A Deep-Learning Approach

Small Unmanned Aircraft Systems (sUAS) are accessible platforms that pose a security threat. These threats warrant affordable and accurate methods for tracking sUAS. We apply neural network-based methods to predict sUAS range from cellphone acoustic recordings; the data comes from twenty-eight cellphones recording three different sUAS that fly over the devices. The timestamped acoustics data is transformed into 0.5s Mel-spectrograms frames and 0.5s raw audio frames. Truth values are calculated using euclidean distance from the sUAS to a cellphone and split into four range classes. The data is sequestered into an 80/20 training-test split and is used to train three different architectures. The 2DCNN architecture outperforms the other architectures (1DCNN and 2DCRNN). The 2DCNN is then re-trained to generalize sUAS range with various sUAS models and achieves an average Macro-F1 score of 0.758 across different sUAS models. The results show that deep-learning-based sUAS ranging with cellphones is an effective and low-cost method for accurately tracking sUAS.