| With the rapid development of information science and technology,indoor localization has received much attentions.Channel state information(CSI)based indoor localization technique has gradually become a research hotspot.In this thesis,CSI amplitude information and phase information are chosen as two modes for deep learning based indoor localization.The main contributions are described as follows:(1)The related theory for CSI localization.The theoretical knowledge of CSI is introduced at first.Then some common machine learning algorithms used in CSI localization system are described.The multi-mode machine learning theory is introduced at last.(2)A CSI amplitude and phase information based localization algorithm using convolutional neural network(CNN)and front-end fusion is proposed.Based on the extracted amplitude information and phase information,the amplitude difference and phase difference based CSI images are constructed to reduce the measurement noise for image construction.Then the Laplacian pyramid fusion algorithm which can keep the original image information as much as possible is used for CSI image fusion.At last,the CNN is used for regression learning and obtain the X-axis coordinate and Y-axis coordinate based regression models.Experimental results demonstrate the efficiency of the proposed algorithm.(3)A two-stream convolutional neural network based localization algorithm using the CSI image fusion is proposed.The amplitude and phase information is used to construct the amplitude difference and phase difference based CSI images at first.Then the CNN is proposed to extract the features of each CSI image respectively.Next,the fused feature of the CSI images is obtained by the linear weight method.Finally,the position regression learning is carried out to obtain X axis and Y axis coordinate based regression models.Experimental results show that the proposed algorithm achieves better localization performance than the existing algorithms. |
PDF Full Text Request