Distance Estimation Based on Deep Convolutional Neural Network Using Ultra-Wideband Signals

Abstract

Recently, high accuracy localization technique is required to provide indoor location services. The purpose of this paper is to propose a distance estimation technique based on deep convolutional neural network (DCNN) for indoor environments. Among distance estimation techniques based on wireless communication signals, the use of ultra-wideband (UWB) signals has the advantage of high accuracy in the time domain. The proposed distance estimation method uses UWB signals and proposes a new DCNN-based distance estimator. The superiority of the proposed method is confirmed through computer simulation. Widely used conventional distance estimators are based on the power threshold. The threshold is determined by signal to noise ratio (SNR) of the received signal. The arrival time of the received signal that exceeds the threshold is considered as the time-of-arrival (ToA) and the distance between transmitter and receiver is obtained from the ToA. On the other hand, the proposed distance estimator requires only the received signal without SNR estimation, which make the proposed technique simpler to implement. According to computer simulation, the conventional method is highly sensitive to SNR and distance. In contrast, the proposed method shows less than 2 m root mean square error (RMSE) performance in a wide range of SNR and the RMSE performance is not degraded in long distances. The proposed distance estimator shows excellent distance estimation performance at low SNR and long distance, so it can be applied to indoor localization system of large indoor space and can be used for precise location service.