基于Hopfield自联想神经网络的天幕靶信号识别技术

为了提高天幕靶系统测试精度和可靠性，探索测试产生各种干扰噪声如弹头激波、弹底激波、蚊虫飞鸟、振动等干扰因素的影响规律，利用Hopfield自联想神经网络的方法，识别并剔除典型因素干扰。通过对实弹射击试验得到的数据进行分析，充分验证了天幕靶系统的准确性和可靠性。分析结果表明：与电平信号识别相比，在射频为5发/min、口径为23 mm的炮弹测试中，自联想神经网络信号识别率提高了17.2%；在弹型为穿甲弹，口径为23 mm的测试中，Hopfield自联想神经网络信号识别率提高了46.7%；对于射频为7 500发/min的天幕靶连发弹丸信号测试条件下，正确信号识别率均达到了93%以上。在复杂环境条件下，Hopfield神经网络算法识别率远远高于传统的电平识别，提高了信号的识别率，能够适应一定区域内的复杂环境因素。

Signal Recognition Technology of Sky Screen Based on Neural Network

The effects of interference factors, such as shock wave of warhead blasting, projectile base shock, aerial birds, insects, vibration, etc.,on sky screen system are analyzed to improve its test accuracy and reliability. The approach of Hopfield auto-associative neural network is used to identify and eliminate the typical interference. The accuracy and reliability of sky screen systemthe is fully validated by analyzing the data from live firing. The results show that, compared with the level signal recognition, Hopfield auto-associative neural network recognition rate can be increased by 17.2% in sky screen test with the 5 bursts in RF/min; Hopfield auto-associative neural network recognition rate is increased by 46.7% in sky screens test of 23 mm caliber armor-piercing shells; under the test condition of firing frequency of 7 500 rounds / minute, the correct signal recognition rate reaches to 93%. In a complex environment, the recognition rate Hopfield neural network algorithm is far higher than traditional level recognition rate，which improve the signal recognition rate and be able to adapt to the complex environmental factors within a region.