Abstract: Among the PDCCH blind detections for the LTE system, a common method used for the acceleration is detecting the valid CCEs in the channel. For the detection of valid CCEs, a common approach is the hard-decision based energy detection combining with fusion rule.However, the reliability of this method will be greatly reduced under low signal-to-noise ratio.To improve the performance, a soft-decision based algorithm for detection of valid CCEs is proposed in this paper. The optimum soft-decision detection threshold satisfying system reliability requirements is deduced by Neyman-Pearson. The probability of false dectection is reduced and the performance of original detection method can be improved by soft-decision. The soft-decision form and threshold for valid CCEs detection are deduced. in this paper. The correctness of the theoretical deduction is proved by simulations. Compared with hard-decision, for instance, the false alarm rate of soft-decision decreases by 82% when the missing alarm rate is 0.15 and SNR = -2dB. The results indicate that the proposed algorithm effectively reduce the missing alarm probability and false alarm probability.