典型的運動感測器包括加速度計、陀螺儀和磁力儀，傳統應用常使用單一類型感測器，然而近年來大量結合多感測器的電子產品快速普及，使得感測器融合成為重要的研究課題。 本論文提出一個智慧型融合方法，根據感測器使用情境和運動特性，利用機率類神經網路(PNN)估算九軸運動感測融合的調節參數，再將九軸資訊融合應用一起，互相彌補感測器不足的部分，如陀螺儀累積誤差的即時校正。我們以三個應用實例來驗證本論文提出的智慧型融合系統和方法的可行性，透過感測器融合系統輸出更穩定且精確的運動資訊。所以，透過九軸運動感測器的智慧型融合系統，使運動量測和辨識結果的精度、解析度、穩定度及回應時間都得到了提升，同時也降低了應用端開發運動感測應用系統的複雜度。Typical motion sensors include accelerometer, gyroscope and magnetometer. In past application we usually use single type sensor. But in recent years, a lot of electronic products are developed rapidly and usually combined with many sensors. Therefore the sensor fusion issue becomes more and more important. In this thesis we proposed the intelligent sensor fusion method regarding the different situations and motion properties. We use PNN to estimate and adjust the parameter of 9-axis motion sensor fusion. And then we can use these information for complementary of other sensor to improve the insufficient part ( ex: real-time calibrating the gyroscope bias drift). In the experiments we can get more stable and accurate motion data by intelligent fusion system. In conclusion, intelligent fusion system of 9-axis motion sensor can increase the system performance, including accuracy, resolution, stability and response time. In the same time it reduces the complexity of developing application system in detecting motion.
