Research Of Error Compensation Technique For Motion Sensor

The on-line handwriting recognition is an important branch in the research field of pattern recognition today. And, in this online handwriting recognition branch, plane handwriting recognition technology has been maturely developed. However, this traditional plane handwriting recognition has its own flaws, it has not a very good user experience. Therefore, in recent years, a novel handwriting recognition has become a new hotspot, that is the space handwriting. Space handwriting recognition gives users much more free space for writing, so it has a better user experience, which would later become the trend of the development of on-line handwriting recognition.The space handwriting equipment need to use acceleration sensors, such sensors are usually MEMS sensors. This is because that the MEMS sensor has the advantages of small size and low cost. Since MEMS technology develops rapidly in recent years, MEMS sensors has achieved the accuracy requirements of use. However, like other sensors, during the process of collect space handwriting acceleration information by MEMS acceleration sensors, the sensor will produce a variety of errors and noise interference. Furthermore, these errors and noise will accumulate over time, causing a large negative impact on the final trajectory reconstruction. Therefore, to eliminate the accumulated error of the acceleration sensor in the space handwriting process is very important and is a subject worthy of study.Through the analysis of some of the existing acceleration sensor error accumulation cancellation technology, the paper proposes two methods to improve. The first is using a gyroscope to improve the recognition rate of the SVM, and finding the exact start and end points of moving by the method of convergence. Using handwriting SVM-based techniques to identify the moving and stationary state surely is better and more adaptive than the threshold value comparison method, but the recognition rate can still be improved. The second approach is to improve the correction of acceleration and speed, the coefficient correction method proposed in this paper can make greater correction where the acceleration error is greater. And make a velocity correction at the turning point in the movement of space handwriting.In this paper, experiments proves the gyroscope can improve the sports and the stationary state recognition rate of the SVM technology of space handwritten. And the convergence of the starting point and end point, the coefficient correction, the turning point in the velocity correction, these three methods can make improvements trajectory reconstruction.