Millimeter-Wave Sparse Channel Estimator Based-on Twin Support Vector Regression in Deep Multipath Environments

Abstract

We develop in this work a Twin Support Vector Regression (TSVR) algorithm built on Discrete Wavelet Transform (DWT) which is operated to 73-Ghz wireless channels in deep multipath indoor environments with -110 dBm receiver sensitivity value. During learning phase, the process builds a denoising procedure based on Discrete Wavelet Transform to avoid outliers and wrong samples. Furthermore, path loss model of Close-In (CI - FSPL) free space reference distance is described and the propagation in large-scale is investigated in terms of (SF) Shadow Factor, probability distribution functions and PLE (Path Loss Exponent) for 73 GHz transmission band. Performance are assessed in terms of BER (Bit Error Rate) and constellation diagram pattern according to 73 GHz frequency, 256-QAM modulation scheme and -110 dBm receiver sensitivity threshold for simple (182 paths) and complex conference rooms (250 paths). The efficiency of the suggested approach in comparison to other conventional techniques has been demonstrated by simulation and experimental results.