Indoor fingerprinting geolocation using wavelet-based features extracted from the channel impulse response in conjunction with an artificial neural network

This paper proposes a method to localize a mobile station in an indoor environment using wavelet-based features (WBF) extracted from the channel impulse response (CIR) in conjunction with an artificial neural network (ANN). The proposed localization system makes use of the fingerprinting technique and employs CIR information as the signature and an artificial neural network as the pattern matching algorithm. For the considered indoor environment, the obtained CIR information can not be applied directly to the input of the ANN due to the high number of the CIR samples since an ANN with a high number of inputs requires a high number of learning patterns during its training. Consequently, relevant features reflecting the CIR signature have to be extracted and then applied to the ANN. The relevant features may be some physical channel parameters or a compressed version of the CIR signature. In this paper, the extraction of the CIR features is done using a wavelet-based compression. The particularity of the method is in the representation of the CIR signature in a judicious way facilitating the design of the ANN. Moreover, when the extracted features correspond to the CIR signature, the localization system tends to give mobile location with a high precision. Simulation of measured CIR in an indoor environment, showed a precision of 2 meters for 91% and 70% of trained and untrained data, respectively.