An energy detection method is provided. The method obtains an initial time point of an input signal with reference to a digital signal corresponding to the input signal. An ith sample value is obtained by sampling the digital signal. The (i−M−N+1)th to the (i−N)th sample values and the (i−N+1)th to the ith sample value are buffered. The weighting coefficient is set to with initial value. A first parameter is obtained according to the weighting coefficient and the (i−M−N+1)th to the (i−N)th sample values, and a second parameter is obtained according to the (i−N+1)th to the ith sample values. A comparison between the first parameter and the second parameter is made, and the sampling time point of the (i−N+1)th sample value is determined as the signal initial time point when the second parameter is greater than the first parameter.