Charrada Anis, Samet Abdelaziz
SERCOM-Labs, EPT Carthage University, 2078, La Marsa, Tunis, Tunisia.
INRS, EMT Center, 800 de la Gauchetière W., Suite 6900, Montreal, QC, H5A 1K6, Canada.
Heliyon. 2020 Nov 9;6(11):e05369. doi: 10.1016/j.heliyon.2020.e05369. eCollection 2020 Nov.
In this article, an effective millimetric wave channel estimation algorithm based on Twin Support Vector Regression (TSVR) is proposed. This algorithm exploits Discrete Wavelet Transform (DWT) in order to denoise samples in learning phase and then enhance fitting performance. An indoor complex conference room environment full of furniture and electronic equipments is adopted for experiments. Through the proposed approach, channel frequency responses are directly estimated using the Orthogonal Frequency Division Multiplexing (OFDM) reference symbol pattern by solving two quadratic programming problems in order to improve generalization aptitude and computational speed. We consider in this work a Channel Impulse Response (CIR) of 60 GHz multipath transmission system generated by the "Wireless InSite" ray tracer by Remcom. The numerical experiments confirm the performance of the proposed approach compared to other conventional algorithms for several configuration scenarios with and without mobility.
本文提出了一种基于双支持向量回归（TSVR）的有效毫米波信道估计算法。该算法利用离散小波变换（DWT）在学习阶段对样本进行去噪，进而提高拟合性能。实验采用了一个充满家具和电子设备的室内复杂会议室环境。通过该方法，通过求解两个二次规划问题，利用正交频分复用（OFDM）参考符号模式直接估计信道频率响应，以提高泛化能力和计算速度。在这项工作中，我们考虑了由Remcom公司的“Wireless InSite”射线追踪器生成的60GHz多径传输系统的信道冲激响应（CIR）。数值实验证实了与其他传统算法相比，该方法在有无移动性的几种配置场景下的性能。
