Abstract-Communication signals are often cyclostationary, that is they have statistical characteristics that vary periodically in time. The cyclic spectrum, a characteristic function of such signals, exhibits spectral peaks at certain locations, called cyclic frequencies. These locations as well as the cyclic spectrum support are defined by the signal parameters, in particular carrier frequency, bandwidth and symbol rate. In this paper, we propose an estimation algorithm that extracts these from the signal cyclic spectrum. This algorithm can be applied to multiband signals, namely signals composed of more than one transmission. Prior to parameter estimation, the number of signals is first estimated. Exploiting the cyclostationarity of communication signals improves the robustness to noise of the parameter estimation. In particular, the proposed algorithm can be used for Cognitive Radios, which traditionally deal with low signal to noise ratios multiband signals, for spectrum sensing purposes by estimating the carrier frequencies and bandwidths. Simulations demonstrate estimation from synthesized and RF Nyquist samples as well as subNyquist samples.