Threshold Selection for Ultra-Wideband TOA Estimation based on Neural Networks

Abstract

Because of the good penetration into many common materials and inherent fine resolution, Ultra-Wideband (UWB) signals are widely used in remote ranging and positioning applications. On the other hand, because of the high sampling rate, coherent Time of Arrival (TOA) estimation algorithms are not practical for low cost, low complexity UWB systems. In order to improve the precision of TOA estimation, an Energy Detection (ED) based non-coherent TOA estimation algorithm using Artificial Neural Networks (ANN) is presented which is based on the skewness after energy detection. The expected values of skewness and kurtosis with respect to the Signal to Noise Ratio (SNR) are investigated. It is shown that the skewness is more suitable for TOA estimation. The best threshold values for different SNRs are investigated and the effects of integration period and channel modes are examined. Comparisons with other ED based algorithms show that in CM1 and CM2 channels, the proposed algorithm provides higher precision and robustness in both high and low SNR environments.