Joint channel and noise variance estimation for OFDM by compressed sensing

In this paper, an iterative and joint algorithm is proposed to estimate additive white Gaussian noise (AWGN) variance and channel impulse response (CIR) in orthogonal frequency division multiplexing (OFDM) systems. With the aid of pilots, a rough CIR is firstly estimated by compressed sensing techniques such as sparsity adaptive matching pursuit (SAMP). By exploiting the sparsity of the estimated CIR and the method of orthogonal projection matrix, the noise subspace is obtained to evaluate noise variance. The above two estimator feeds each other, then the estimation of noise variance is used to update the CIR estimation which also updates the noise variance in the following. After several iterations, a perfect performance can be achieved. The operation process is simple and its performance is robust to the noise power. Simulation results show that the propose scheme can obtain better performance than traditional scheme with less pilots.