基于EMD与分批估计的动态称量快速融合方法

Fast fusion method of dynamic weighing based on EMD and batch estimation

活体动物动态称量在生物医学实验中具有举足轻重的作用。为实现活体动物快速、准确的称量,提出基于EMD与分批估计相结合的动态称量信息融合方法:首先利用经验模态分解(EMD)方法解析活体动物动态称量的采样信号,提取表征采样信号的趋势或均值的残余量,计算各阶本征模函数(IMF)与采样信号的相关系数,根据相关系数分辨虚假IMF,并将虚假IMF和原残余量之和作为新残余量;然后对新残余量进行分批估计,消除各种噪声、扰动对称量结果的影响,获得准确的动态称量值。通过建立动态称量系统的仿真模型,对基于EMD与分批估计相结合的动态称量信息融合方法进行了仿真实验,仿真结果表明该算法准确度高。为了验证基于EMD与分批估计相结合的动态称量方法的有效性,进行标准砝码的静态称量和小鼠的动态称量试验,实测结果表明该方法稳定、可靠,称量误差≤±0.5%。

Living animal dynamic weighing plays a decisive role in biomedical experiment. In order to weigh a living animal fast and accurately,a fast fusion method of dynamic weighing based on EMD and batch estimation is presented in this paper. Firstly,the Empirical Mode Decomposition（ EMD） method is used to analyze the sampled signal of living animal dynamic weighing,extract the residual amount describing the trend or mean of the sampled signal; and the correlation coefficient of each order intrinsic mode function（ IMF）and the sampled signal is calculated. The false IMFs are distinguished according to the correlation coefficients,the summation of the false IMF and the original residual amount are taken as the new residual amount. The batch estimation is performed on the new residual amount to eliminate the effects of different kinds of noises and disturbances on the weighing results,and the accurate dynamic weighing value is obtained. The simulation model of the dynamic weighing system was established,and simulation experiments were carried out on the information fusion method of dynamic weighing based on the combination of EMD and batch estimation. The results of the simulation experiments show that the algorithm achieves high accuracy. In order to verify the effectiveness of the fast fusion method of dynamic weighing based on EMD and batch estimation,the static weighing test of the standard weights and dynamic weighing test of mice were conducted. The experiment results show that the method proposed in this paper is stable and reliable,and the weighing error is less than 0.5%.