【摘要】 由于测量噪声的存在且噪声标准差未知,采用最小二乘法(LS)无法得到传感器动态补偿器参数的无偏估计.研究了一种有输入噪声扰动的无偏参数辨识算法,先通过小波变换估计输入信号噪声的方差,再由估计得到的方差,通过偏差消除的递推最小二乘法,对补偿器的参数进行无偏辨识.该算法降低了对输入辨识信号为白噪声的要求,具有较强的实用性,且因采用了递推运算,可以用于传感器动态补偿器的在线辨识.仿真实验验证了该方法的有效性.更多还原

【Abstract】 Because the sensor output data are contaminated with noise and the variance is unknown,when the ordinary least squares method is applied to the parameter estimation of the sensor’s dynamic compensator with noisy data matrix,the estimates turn out to be biased.This paper proposed a new identification algorithm for unbiased parameter estimation.To implement this algorithm,the variance of sensor output data sequence is estimated using the wavelet transform,and the bias-eliminated recursive least-squares is used to estimate the parameter of compensator.But the algorithm does not require the input signals to be white noise with zero mean,carries out the recursive computation of bias eliminated and can be on-line implemented.The experimental results show that the approach is effective.更多还原