基于相关向量机的MEMS加速度计零偏温漂补偿

应用相关向量机（RVM，Relevance Vector Machine）回归预测方法建立了基于RVM的零偏温漂预测补偿模型，利用温度值和温升速率作为输入可预测不同温变过程下的加速度计零偏温漂，探讨了不同训练样本数量、核函数类型和不同的核函数宽度对预测补偿效果的影响，最后应用环境温度试验数据对模型进行检验和验证.结果表明：基于RVM的零偏温漂预测补偿模型能够有效的预测微机械（MEMS，Micro Electro Mechanical Systems）加速度计零偏温漂，模型预测的均方根误差小于1%，补偿后滞环误差最大由0.06g缩减为0.015g.

Thermal bias drift compensation of MEMS accelerometer based on relevance vector machine

Thermal bias drift prognosis and compensation model was built based on the regression algorithm of relevance vector machine. The thermal bias drift of the accelerometer experiencing different temperature load can be classified by using both the temperature and the temperature rate as the model input. The influence of training sample number, the kernel function and the parameter sigma were discussed. Experimentation with the data of the temperature cycling test was conducted. According to the experimental result, the thermal bias drift of the accelerometer can be prognosed accurately by the model, the mean square error is less than 1%, and the size of the thermal hysteresis loop is reduced from 0.06g to 0.015g.