基于反相双峰指数模型的微弱瞬态极低频信号的估计与检测

Estimation and detection of the weak transient ELF signal based on the phase inverting double-peak exponential model

在铁磁管道环境下,同频窄带噪声中的微弱瞬态极低频信号的估计与检测问题是管道机器人跟踪定位中最为重要的科学问题之一。为实现窄带噪声中的微弱瞬态信号的有效检测,分析了接收线圈在不同空间姿态下的信号特点,针对信号与窄带噪声相位上的区别以及信号包络的形态,建立了相位反相的双峰指数函数数学模型;应用非线性最小二乘估计,实现了该数学模型和真实信号之间的拟合;通过仿真和实验验证了该数学模型与真实信号的高度匹配性,并应用蒙特卡洛仿真分析了该模型参数估计的性能;使用所建立的数学模型,构建了平均功率检测器和瞬时最大功率检测器,通过分析比较极低频发射机在不同移动速度和接收信噪比条件下的检测性能,得出了两种检测器的特点和适用范围,并指出在现有工程背景下平均功率检测器的优势。实验证明,相位反相信息对同频窄带噪声中的信号检测非常重要,平均功率检测器在低信噪比条件下的检测性能良好。在信噪比为0.05 d B、虚警概率设为1%时,该检测器的检测概率达98.8%。

It is an important problem in the localization system of pipeline robot that the estimation and detection of weak transient ex- tremely low frequency （ELF） signal with the same frequency narrowband noise. In order to efficiently detect the signal, the characteris- tic of the received signal is analyzed when the coil transducer is in different space posture, and mathematic model is built with a double- peak exponential function with phase jump by distinguishing the phase＇ s characteristic between the signal and the noise in same frequen- cy narrowband, as well as the envelope of the signal. The good similarity between the real signal and the mathematic model is proved in the simulation and the experiment. Then, according to nonlinear least square estimation, the matching between the real signal and the mathematic model is realized, and the performance of estimation is analyzed by Monte-Carlo simulation. Based on the estimated math- ematic model, a mean power detector and a maximum power detector are built respectively. Their feature and circumstances of applica- tion are given by analyzing their performance at different speed of the ELF transmitter and background noise. It is proved that the mean power detector is much better than the maximum power detector under the current engineering conditions. The experimental results show that the phase jump is important to distinguish the signal and the noise as well as to guarantee the good performance at low Signal to Noise Ratio （SNR） for the mean power detector. The detection probability of this detector could be as high as 98.8% under the conditions in which the SNR is 0.05 dB and false alarm probability is 1%.