Effective detection problem for BPSK (binary phase shift keying) signals under condition of Gaussian white noise environments is studied. In cognitive radio, when second user senses spectrum from first user, this makes the best detection threshold with the minimum of sum of false alarm and missed detection rate in existence. By theoretical derivation for distribution character of energy detection static, the best detection threshold problem is converted to solution of the minimum of the sum, computation expression is determined by sampling length, noise variance and signal to noise ratio (SNR) are given. The simulation of sum value curve variety in different SNR shows that the best detection threshold is in existence. The performance of the best threshold is even better than before by analysis for detection ratio performance between the threshold and the one in constant false alarm ratio (CFAR), false alarm ratio performance between the threshold and the one in constant detection ratio (CDR), other two thresholds' convergence to the threshold performance. Besides, Monte Carlo experiment shows that the mathematic analysis is correct.