| In modern warfare,surprise attacks on important targets by ultra-low-flying drones,helicopters,fighter jets,and cruise missiles have become the main combat mode.However,the radar system,which is mainly responsible for air monitoring,has blind spots for ultra-low-altitude flying targets,and the monitoring of ultra-low-altitude flying targets has become an important research subject for air defense safety.In order to improve the monitoring ability of targets coming from ultra-low altitude flight,this paper designs a set of low altitude target recognition system based on passive acoustics,which is supported by the project of "Research on low altitude target passive acoustic detection and recognition technology based on voiceprint feature"(19-H863-01-ZT-002-086-01),which is cooperated by 202 Institute and Chang ’an University.The main research contents include: discussing the theory required by the system and designing the overall plan of the system in combination with the subject index requirements;In order to improve the signal-to-noise ratio of the signal,the signal conditioning circuit is designed,and the power supply circuit,data acquisition and data transmission circuit are designed at the same time;In order to improve the recognition rate of the system,a low-altitude target composite neural network model is proposed,and the software design of data acquisition,data transmission,acoustic signal feature extraction,and low-altitude target recognition is completed and a data set is established.Finally,an experimental platform is built to verify and analyze the whole system.Through the experiment of the influence of the signal conditioning circuit board on the recognition rate,the recognition rate of the system is improved after the signal conditioning circuit board is connected.Through experiments on different neural network models,the results show that the improved composite neural network model has a greater improvement in recognition rate.The actual test and simulation test results show that the low-altitude target recognition system designed in this paper achieves a recognition accuracy higher than92.875% within the effective monitoring range. |
