MMSE Detection Based on Noise Statistics with Random Noise Variance

In most conventional signal detection techniques, the noise is assumed to be Gaussian distributed with known variance at the receiver. In this paper, a novel detector using a soft quantization technique based on limited knowledge of the noise statistics is proposed. A promising application of the novel detector is the data detection in orthogonal frequency division multiplexing (OFDM) systems aiming at minimum mean square error (MMSE). Noise on each subcarrier is assumed to be Gaussian distributed with random variance having a certain probability density function (PDF). The distribution of the variance is derived from various realistic scenarios. Assuming that the noise variance is the same in a cluster of OFDM subcarriers, the proposed soft quantizer exploiting the PDF of the noise variance can achieve improved system performance as compared to conventional soft quantizers employing a separate noise variance estimator. Furthermore, log-likelihood ratios (LLRs) are obtained for promising applications of this soft quantizer.