Q-Flex Accelerometer Error Correction Based on the Support Vector Method

The algorithm for error correction in the Q-flex accelerometer signal based on the support vector machine (SVM) method is proposed. This method uses the temperature and accelerometer output measured in real time as input parameters. The algorithm is characterized by simple implementation and high speed of processing on computing equipment due to the elegance of the method in linear and nonlinear regression problems. Regularization in SVM eliminates the multicollinearity and reduces the set of coefficients in overfitting. Regularization is equivalent to introducing a priori distribution in the space of coefficients of the approximating model. The results of the compensation of the model obtained by the SVM are presented.