基于3DCNN的CSI-cluster室内指纹定位算法

针对室内环境中复杂的多径效应影响定位精度问题，提出一种基于3维卷积神经网络(3 dimensional convolutional neural network，3DCNN)多径程度划分的自校准指纹定位算法。该算法利用MeanShift方法分析定位区域内每一个采样点的信道状态信息数据分布特性，得到其可代表多径效应程度的簇类数量，结合阈值原则将指纹库划分为2种不同多径程度的子库，从而减少多径程度差异较大的指纹点对后续定位影响利用3DCNN深度学习2类指纹子库。在定位阶段，根据校准算法判断待测数据所属子库，并采用相应的3DCNN模型估计位置。通过仿真实验验证，该方法在保证指纹库构建合理性和高效性的同时，在定位精度方面实现了明显的提升，优于与之对比的相关算法。

CSI-cluster indoor fingerprint localization algorithm based on 3DCNN

Aiming at the problem of location accuracy in the multi-path effect of indoor environment, this paper proposes a multi-path degree division with self-calibration fingerprint localization algorithm based on 3 dimensional convolutional neural network (3DCNN). The algorithm uses the MeanShift method to analyze the data distribution characteristics of the channel state information of each sampling point in the location area, and obtains the number of clusters that can represent the degree of multipath effect.The fingerprint library is divided into two sub-libraries with different multipath degrees by combining the threshold principle, thereby reducing the influence of fingerprint points with large differences in multipath degree on subsequent positioning, and using 3DCNN to deeply learn two types of fingerprint sub-libraries;in the positioning stage,judging the sub-library to which the data to be tested belongs according to the calibration algorithm, and adopting corresponding 3DCNN model estimates the location. The simulation experiment proves that this method can achieve a significant improvement in positioning accuracy while ensuring the rationality and high efficiency of the fingerprint database construction. This method is better than the related algorithms.