基于半监督极限学习机的增量式定位算法
发布时间：2018-12-15 06:39
【摘要】：针对目前室内指纹定位算法存在实时性差、对动态环境适应性不足的问题,提出一种新的基于半监督极限学习机的定位算法。该算法首先通过半监督极限学习机建立初始化位置估计模型,然后利用新增的半标记数据对原定位模型进行动态调整,最后为新增训练数据分配合适惩罚权重,使模型具有时效机制。仿真结果表明,该定位算法在保证定位实时性的同时提高了对动态环境的适应性。
[Abstract]:A new localization algorithm based on semi-supervised extreme learning machine is proposed to solve the problem that the indoor fingerprint localization algorithm is poor in real time and adaptability to the dynamic environment. The algorithm first establishes the initialization position estimation model by semi-supervised extreme learning machine, then dynamically adjusts the original location model by using the newly added half-mark data, and finally assigns appropriate penalty weight to the new training data. The model has a time-effect mechanism. The simulation results show that the algorithm not only ensures the real time of location, but also improves the adaptability to dynamic environment.
【作者单位】： 江南大学物联网工程学院;
【基金】：江苏省产学研联合创新资金前瞻性联合研究项目(BY2014023-31) 江苏省“六大人才高峰”项目(WLW-007)
【分类号】：TN929.5;TP181;TP212.9
本文编号：2380151
[Abstract]:A new localization algorithm based on semi-supervised extreme learning machine is proposed to solve the problem that the indoor fingerprint localization algorithm is poor in real time and adaptability to the dynamic environment. The algorithm first establishes the initialization position estimation model by semi-supervised extreme learning machine, then dynamically adjusts the original location model by using the newly added half-mark data, and finally assigns appropriate penalty weight to the new training data. The model has a time-effect mechanism. The simulation results show that the algorithm not only ensures the real time of location, but also improves the adaptability to dynamic environment.
【作者单位】： 江南大学物联网工程学院;
【基金】：江苏省产学研联合创新资金前瞻性联合研究项目(BY2014023-31) 江苏省“六大人才高峰”项目(WLW-007)
【分类号】：TN929.5;TP181;TP212.9
【相似文献】
相关会议论文 前1条
1 单莘;;一种网络告警的增量式情景规则挖掘方法[A];中国通信学会第五届学术年会论文集[C];2008年
，本文编号：2380151
本文链接：https://www.wllwen.com/kejilunwen/xinxigongchenglunwen/2380151.html
