Acoustic Target Recognition Method Based on EEG Signals

In this paper, a novel acoustic target recognition method is proposed by decoding the Electroencephalogram (EEG) signals of an operator when perceiving environmental sound. Taking unmanned aerial vehicle (UAV) detection (detection of the presence of drones from sound) as an example, we recorded real environment noise and target sound. Then the experimental paradigm was designed to simulate real acoustic target detection. Clear event-related potentials (ERP) were observed from the EEG signals of 4 subjects. We extracted the time domain features of the EEG signals based on the observed neural representations and designed a CNN(convolution neural network)-based classifier to distinguish the EEG signals in two different states ("normal "versus" target") which was compared with the traditional SVM(support vector machine)-based classifier. The results show that the classification accuracy based on CNN reaches 81.25%, higher than SVM. The method proposed in this paper can be used as the theoretical basis for adding human intelligence to perceive the environment in a target detection system.