Please clarify Your question. for example, give more details on: do You suppose noise is white and gaussian? do you simultaneously have signal? If yes, what is expected signal-to-noise ratio and what is a priori information on signal properties? do you assume that You have noise (or mixture of signal and noise) in digital form?
There can be several advices depending upon Your answers to my questions. For example, I mean that Your task has more solutions of received signal has been already transformed to digital form. Second, it's easier to estimate noise power (variance) if there are intervals of time when a signal is absent and you are able to detect such intervals. Third, Your estimate can be more accurate if noise pdf is close to Gaussian. If it has heavy tail distribution, then accuracte estimation is more problematic.