论文部分内容阅读
针对动态Allan方差运用固定长度的分析窗截取信号导致样本数据量减少,长相关时间下方差估计值置信度降低,首先,针对动态信号跟踪能力与置信度的提高不能兼顾的问题提出了一种改进算法。引入截断窗内峭度值作为表征信号短时稳定度的参数,并建立以峭度为变量的窗宽函数,该函数可以使截断窗长随着信号的平稳程度自动变化。其次,再用窗宽自适应的滑动窗分段截取陀螺随机误差,分别对每个截断窗内样本进行总方差计算以增加方差估计的自由度。最后,计算延伸后样本的Allan方差,并将其以三维形式排列出来。结果表明:应用该方法对光纤陀螺启动信号进行分析,该算法既能更有效地跟踪信号的非平稳变化,又能大幅降低长相关时间下方差的估计误差。
Aiming at the problem that the dynamic Allan variance intercepts signals with a fixed length of analysis window, the amount of sample data is reduced and the confidence of the variance estimation under long correlated time is reduced. At first, an improvement is proposed to solve the problem that dynamic signal tracking ability and confidence can not be improved algorithm. The kurtosis value in the truncation window is used as the parameter to characterize the short-time stability of the signal, and the window width function is constructed with kurtosis as the variable. This function can make the truncation window length change automatically with the signal smoothness. Secondly, the gyro random error is intercepted by the sliding window segment with adaptive window width, and the total variance of the samples in each truncated window is calculated to increase the degree of freedom of variance estimation. Finally, the Allan variance of the extended sample is calculated and arranged in three dimensions. The results show that this method can be used to analyze the start-up signal of fiber optic gyroscope (FOG). This algorithm not only can effectively track the non-stationary signal changes, but also can significantly reduce the estimation error of variance under long correlated time.