Data-classification-based SNR estimation for linearly modulated signals

We present a new numerical approach to SNR estimation of linearly modulated signals after passing through a complex additive white Gaussian noise (AWGN) channel. This classified data (CD) based SNR estimator is particularly suitable for both constant and non-constant modulus constellations, including BPSK, M-PSK and M-QAM. In essence, the received data will be first classified into a number of classes, and then a look-up table (LUT) is searched to find an entry that closest matches with the classified data; this matched result in LUT corresponds to the SNR value of the received data. The performance of the proposed estimator in terms of accuracy and complexity is evaluated by numerical simulations and compared with a few well-known estimators such as moment and maximum likelihood based estimators.