Hardware Impairment Estimation in NB-IoT: A Parallel Multitask Learning Method

Abstract

Orthogonal frequency-division multiplexing (OFDM) is widely adopted in narrowband Internet of Things (NB-IoT). Nevertheless, the OFDM system is highly sensitive to the impairments caused by imperfect radio-frequency hardwares, which may greatly jeopardize the orthogonality between different subcarriers and degrade the demodulation performance. Though large efforts have been devoted to the hardware impairment estimation, however, it is highly challenging to jointly estimate multiple hardware impairments due to their coupling effects, especially for the NB-IoT systems which usually work in low signal-to-noise ratio (SNR) regions with limited computing and radio resources. In this article, we propose a parallel multitask learning (MTL) estimator to jointly estimate carrier frequency offset and in- and quadrature-phase imbalance in NB-IoT systems. Specifically, MTL is introduced to extract the inherent correlations between different hardware impairments so as to address the coupling effect and average the noise impact. In addition, we propose a parallel structure and a sliding window scheme to reduce the network complexity and decrease the estimation bias. Numerical results show that our proposed parallel MTL estimator can jointly estimate multiple hardware impairments with short pilot sequences, and outperform the conventional methods in terms of estimation accuracy and computation time in the typical SNR regions of the IoT devices.