The demand for radio spectrum is rapidly increasing for applications such as mobile telephony, digital video broadcasting (DVB), wireless local area networks (WiFi), and wireless sensor networks (ZigBee), and internet of things. Indeed, these resources are becoming increasingly scarce or even nonexistent. This scarcity has led to the concept of Cognitive Radio (CR) communication which has used to scarce and limits natural resources efficiently without any interference to the primary users (PUs). Among most challenging problems in cognitive radio systems is spectrum sensing concepts. Many strategies exist in the literature for satisfactory detection performance. Energy detection (ED) is popularly preferred approach due to its implementation simplicity and low complexity. But, ED model based on imperfect knowledge of noise power which is difficult to obtain under low SNR environment. Furthermore, sensitivity to noise power fluctuations. In this context, this work proposes and evaluates an improved version of the energy detection algorithm calls Mean ED (MEED) that is able to outperform the classical energy detection scheme. This new technique is inspired of oversampling and bootstrap methods. The performance improvement is evaluated analytically and corroborated with the simulation results.