Research Of Accumulated Error Elimination Technique For Accelerometer In Space Writing

Space handwriting is a novel HCI technology and offers unrestrained writing space for users compared with the traditional plane handwriting input. With the development of MEMS technology, small-sized and low-cost MEMS accelerometers are extensively adopted as the handwriting devices. However, all kinds of errors and noises accumulate during the writing procedure, causing an adverse impact on the result. Hence, accumulated error elimination targeted for accelerometers is critical.By analyzing the existing error elimination techniques, a SVM-based error elimination technique is proposed to overcome two shortcomings. Aiming at the poor adaptability of threshold-based methods, SVM is utilized to discriminate motion and non-motion states. To tackle the problem that trajectory reconstruction is not good due to correcting velocity only, a solution correcting acceleration and velocity both is presented. Based on these two points the research moves on to successive characters input in space. The discrimination between two states is extended to three states, namely effective writing, stillness and pen-up-down, where pen-up-down state works as the separator between successively input characters. The correction of acceleration and velocity is also revised to adapt to trajectory reconstruction of multi-segments.In the end, experiments are conducted to verify the effectiveness of SVM-based error elimination technique and successive characters input. It is first proved that the method that amends acceleration before velocity gains better trajectory reconstruction effect. Then in comparison with the MAD method, it comes out that state discrimination based on SVM not only guarantees good trajectory reconstruction but also possesses superior adaptability. While trajectory separation between successive characters is successfully achieved with the introduction of pen-up-down state.