CSI Indoor Location Technology Based On Image Features

With the development of wireless technology and machine learning theory,indoor localization using radio signals and machine learning has received much attention.Due to the popularity of network infrastructure and high coverage of Wi-Fi signals,in this paper,we study indoor localization technology based on Wi-Fi signals.Specially,choosing the channel state information(CSI)in the Wi-Fi signal as the localization parameter,machine learning method is proposed for position estimation.The main work of the thesis includes:(1)Related theoretical knowledge based on CSI indoor localization is studied.Firstly,the fingerprint based localization model using CSI information is introduced.Then,some existing machine learning algorithms of CSI localization are described.At last,the image feature extraction methods are proposed which provides a solid theoretical foundation for the following research.(2)A gray-level co-occurrence matrix and factor analysis based indoor localization algorithm using a support vector machine(SVM)is proposed.First,the CSI amplitude image is constructed using the CSI amplitude information.Then,the gray-level co-occurrence matrix is used to extract the texture features of the CSI image.Next,the factor analysis dimensionality reduction method which can extract the main information of the data is used for feature dimension reduction.Finally,the SVM is used for location-based classification learning.The position classification model is obtained.The proposed algorithm improves offline classification learning performance through exploiting the spatial correlation properties of gray levels in images.The experimental results verify the efficiency of the proposed algorithm.(3)A gray level co-occurrence matrix,wavelet transform and t-SNE dimensionality reduction based indoor localization algorithm by SVM is proposed.Firstly,the CSI information is used to construct the CSI amplitude image.And then the gray-level co-occurrence matrix and wavelet transform are used to extract image texture features.Next,the t-SNE dimensionality reduction is used to remove redundant information and extract the main information of the data.Finally,the SVM is used for classification learning and to obtain the position classification model.The algorithm improves the offline classification learning performance by extracting image features by exploiting the spatial correlation properties of gray levels and the multi-scale characteristics of the wavelet transform.Experimental results show that the proposed algorithm has better localization performance than existing algorithms.