基於低取樣頻率慣性感測器之行人軌跡系統

Abstract

在行動裝置和現代人生活越來越密切的趨勢下，如何延長行動裝置電源之使用壽命是一個很熱門的議題，經過實際量測確認降低慣性感測器頻率可以有效減少能源的消耗。因此本篇論文提出『基於低取樣頻率慣性感測器之行人軌跡系統』，此系統以特徵值萃取加速度計投影過後的波形，並藉由統計一般大眾行走數據以設定正常行走時所產生波形的參數上下限值。依照多數決的投票方式來判定步行動作，以計算出行人移動步數。利用實驗所得的步距二次迴歸方程式，推測行人一步移動的距離。最後使用電子羅盤計算行人的行走方向。 此系統利用特徵值萃取的特性，成功地降低慣性感測器的取樣頻率，並維持接近的定位精準度。經由實際量測確認10Hz取樣頻率能比40Hz取樣頻率，節省55%的能耗。因此本系統提出的軌跡定位技術，非常適合用來發展高精確的個人導覽系統並且整合在能源有限的行動裝置設備上。

Prolonging battery lifetime of mobile devices is a hot study in recent years. According to experiment measurement, it can reduce power consumption of mobile devices by lowering the sampling rate of Inertial Measurement Unit (IMU). This thesis proposes a “Low Sampling Rate IMU-based Pedestrian Trajectory System (LSR-PTS)” which extracts waveform features of vertical acceleration to come out the upper and lower bound parameters of stride waveform by statistical experiment result. By using voting-based algorithm to verify stride waveform to get stride count. A quadratic regression analysis equation of stride length is used to predict the stride length. Finally, the walking direction is calculated by digital compass of smart phones. We proposed LSR-PTS which uses features-based extraction method to keep similar positioning accuracy and reduce power consumption by lowering sampling rate of IMU. According to experiments, IMU with 10Hz sampling rate can reduce 55% power consumption than IMU with 40Hz sampling rate. Therefore, this system is suitable to develop the high-precision personal navigation system with power constraint on mobile devices.