Noise energy estimation is widely used as a pre-process in speech enhancement and speech recognition systems. While many signal processing algorithms have been proposed to estimate the additive noise energy, they are generally based on some statistical hypothesis and have high computation complexity, which is crucial in mobile devices. When the hypothesis does not hold, the estimation performance is not good enough. In this work, we present a real-time low complexity noise estimation algorithm for speech enhancement. We adopt a new strategy to estimate noise energy in adverse environments. Compared with traditional methods, the proposed method is simple and easy to tune. Experiment results show that the proposed method can provide reasonable estimation results.