机构地区: 北京邮电大学软件学院
出 处: 《通信学报》 2010年第11期195-204,共10页
摘 要: 基于贝叶斯滤波框架,提出了基于核函数法及马尔可夫链的节点定位算法,该算法采用射频指纹匹配技术,使用核函数构建似然函数,充分利用观测与多个训练样本之间的相似性,避免使用先验确定型信号分布模型产生的误差。此外,为提高移动目标的定位精度和定位实时性,该算法还使用马尔可夫链,通过利用目标的历史状态和环境布局等信息对匹配定位的网格搜索空间进行限制,剔除目标移动过程中不可能发生的位置跳变。实验证明,与高斯分布模型相比,所提定位算法具有更高的定位正确率和定位精度。 To position indoor objects accurately and robustly,a novel node localization based on kernel function and Markov chains was presented,which employs Bayesian filter framework and radio fingerprinting technology.It uses kernel function to construct likelihood function to take full advantage of the similarity between observation and several training samples,which avoids the error brought by employing a priori determined distribution model.Furthermore,the proposed algorithm uses Markov chains to improve the localization accuracy and shorten the positioning time.It limits the search space of the matching grids with object's previous state and the environment layout,and refuses the object’s impossible position jump during the moving process.Experiments confirm that the proposed localization outperforms the algorithm with Gaussian distribution model.
领 域: [自动化与计算机技术] [自动化与计算机技术]