Almost Pilotless Channel Estimation Using Dimension Reduction by PCA
Hanho Wang · 2023
In this paper, an almost pilotless channel estimator (APCE) using machine learning based on Principle Components Analysis (PCA) is proposed. The APCE extracts PCA vectors from a wireless communication channel and estimates the phase and amplitude of the channel using an ML model trained on PCA coefficients. Through the highly accurate phase estimation using the ML and the amplitude estimation process that suppresses noise, the APCE obtains an signal-to-noise ratio (SNR) gain of 6 to 7dB compared to the least-squared channel estimator in the SNR region where QPSK operates. Research results that can reduce the computational complexity that inevitably occurs while using ML are also provided at the end of the paper.