脉冲噪声下微弱信号相似度检测方法及应用

实际中采集的信号常常含有脉冲性噪声且信噪比偏低,这时用相关函数来表示信号之间的相似度时,会出现精度不高、鲁棒性不强的问题。提出了一种脉冲噪声下微弱信号的相似度检测方法,即非线性变换法。该方法采用Alpha稳定分布描述脉冲噪声信号,首先将采集的带噪信号进行Sigmoid映射,削弱脉冲噪声对信号相似度检测的干扰;接着对映射后的信号求取相关函数,得到信号之间相似度的信息。该方法简洁、计算量小,没有调整参数,能较好地减弱脉冲噪声对信号相似度估计的不利影响。该方法可以应用于定位系统中信号到达时差的估计。比较脉冲噪声低信噪比条件下计算机仿真结果显示,非线性变换相关法的估计精度高于共变法。实验结果表明:非线性变换相关法不仅可以用于高斯噪声环境下微弱信号相似度的检测,也可以用于强脉冲噪声环境下信号相似度的检测,具有较宽的适用范围和较好的鲁棒性。

Method and Application of Similarity Detection for Weak Signals under Impulse Noise

In practice, collected signals often contain impulse noise, and have low signal-to-noise ratio. In this case, when the correlation function is use to represent the similarity between signals, problems of low accuracy and robustness will occur. In terms of these problems, a similarity detection method for weaker signals with impulse noise is proposed, i.e., the nonlinear transformation correlation method. In this method, the Alpha stable distribution is adopted to describe the signals with impulse noise. First, Sigmoid mapping is performed on the collected signals with noise to reduce the interference of the pulse noise to the signal similarity detection. Then, the correlation function of the mapped signals is used to obtain the similarity information between the signals. This method is simple, requires a little computation, does not need to adjust parameters, and can effectively reduce the effect of impulse noise on the signal similarity estimation. The method can be applied to the estimation of time different of arrival (TDOA) for signals in position system. With the comparison of the simulation results of signals at low signal-to-noise ratio with impulse noise, it is known that the estimation accuracy of the nonlinear transformation correlation method is higher than that of the co-variation method. The experimental results show that the nonlinear transform correlation method can be used to detect the weak signal similarity not only in the Gaussian noise environment but also in strong impulse noise environment. It has a wide range of applications and good robustness.