The Snr Estimation Method

Parameter estimation is an important part of signal processing. It has been developed fast and well. SNR (Signal-to-Noise Ratio) is an important parameter in many communication receivers, and their proper operations are strongly dependent on the correct estimation of SNR. For instance, SNR estimation is used in power control, mobile assisted hand-off, channel assignment and so on. In addition, various algorithms require knowledge of the SNR for optimal performance. The search for a good SNR estimation technique is motivated by the fast developed signal processing. However, decreasing hardware costs and increasing demands for pushing system performance to the achievable limits make an investigation of SNR estimation techniques timely. This dissertation will be focused on the SNR estimation under AWGN (Additive White Gauss Noise) channels, colored-noise environment and SNR estimation for signal which can be model by AR(Auto-Regressive) process.First of all, the research status of SNR estimation is summarized based on the study of plenty of related literatures. The definition of SNR and the criterion of performance evaluation and for SNR estimation methods are described.Then, the principles and characteristics of recent representative SNR estimation algorithms under AWGN channels are discussed and presented, including time-domainand frequency-domain algotithms, such as ML(Maximum Likelihood) algorithm, M₂M₄ (Second-and-Fourth order Moment) algorithm and so on.Furthermore, a novel SNR estimation method based on two-channel model for colored-noise environments is proposed. We use the phase information from the two-channel model to obtain the signal bandwidth. Hence, the SNR can be estimated. The simulation shows that the novel method performs better than other methods for colored-noise environment.At last, SNR estimation for the signal which can be modeled by AR process is studied in this dissertation. This algorithm takes advantage of the AR model information of the received signal. The classical time-domain methods are applicable for the input signals with certain envelopes. As we known, AR model process has no certain envelope. Compared with the frequency-domain based method, the simulation result shows that the parametric method is more precise than the frequency-domain method.