Abstract: In the wireless indoor positioning research, it proposes using the phone built inertial acceleration sensor data to calculate the user’s motion path, automatically acquires the fingerprint data when passing through the fingerprint, automatically builds fingerprint database to reduce the large amount of data collection in the stage of off-line fingerprint database establishment. Due to cellphone’s built-in acceleration sensors collect data exists non-linear noise and other interference, it proposes an improved GPF（Gaussian Particle Filter）algorithm, the improved filtering algorithm acceleration filtered samples get a better valuation to calculate the motion path. It results a significant effect on state variables obeys linear variation and observed equation obeys nonlinear changes system model. Experimental results show that the improved algorithm to calculate the path is better than Gaussian particle filter algorithm and Kalman filter algorithm and reduces the computational complexity. After the new method to establish the database, it will save a lot of off-line data collection job, but basically the same accuracy as compared with the traditional method of fingerprint database.

Key words: WiFi indoor positioning, inertial sensors, Gaussian particle filter, track, fingerprint database

摘要： 在无线室内定位研究中，为减少离线指纹数据库建立阶段的大量数据采集工作，提出利用手机内置惯性传感器加速度数据，计算用户运动路径，在经过指纹点时自动采集该指纹点的数据，自动动态建立指纹库的方法。针对手机内置传感器采集的加速度数据存在非线性噪声等干扰，提出了改进的高斯粒子滤波GPF（Gaussian Particle Filter）算法，将加速度样本经改进滤波算法获得较好的估值后，再进行运动路径计算。对状态变量服从线性变化，观测方程为非线性变化的系统模型有显著效果。实验结果表明，经改进的算法计算出的路径效果优于高斯粒子滤波算法和卡尔曼滤波算法，并且减少了计算复杂度。新方法建立数据库后，节约了大量的离线数据采集工作，但是精度与传统指纹库方法相比基本一致。

关键词: WiFi室内定位, 惯性传感器, 高斯粒子滤波, 轨迹, 指纹数据库