论文部分内容阅读
本文中研究一种水下目标识别方法并介绍一组材料、大小、厚度不同的中空圆柱形水下目标的识别实验。我们分三步提取水下目标特征量。第一步是在频域上用短脉冲作为入射信号辨识脉冲响应函数。第二步用“卷积-修正递推最小二乘”法估计目标ARMA模型参数。第三步计算频域上的极点。我们用极点作为水下目标的特征量,并认为它是一个不变量。计算机模拟结果表明,同以AR参数作为特征量的方法相比较,极点法的抗噪声干扰能力改善12dB。我们采用匹配滤波分类器对水下目标进行分类。实验结果表明,在无噪声情况下,正确识别率可达100％。信噪比不低于8dB时,正确识别率仍然达到100％。当信噪比是5dB时正确识别率可达87.5%。
In this paper, a underwater target recognition method is studied and a set of experiments for identifying hollow cylindrical underwater targets with different materials, sizes and thicknesses are introduced. We extract underwater target features in three steps. The first step is to use a short pulse in the frequency domain as the incident signal to identify the impulse response function. The second step is to estimate the target ARMA model parameters using the Convolution-Modified Recursive Least Squares method. The third step is to calculate the pole in the frequency domain. We use the pole as a feature of the underwater target and consider it an invariant. Computer simulation results show that compared with the AR parameter as a feature of the method compared to the pole method of anti-noise interference ability to improve 12dB. We use a matched filter classifier to classify underwater targets. Experimental results show that under the condition of no noise, the correct recognition rate can reach 100%. When SNR is no less than 8dB, the correct recognition rate still reaches 100%. When the SNR is 5dB, the correct recognition rate can reach 87.5%.