In this paper, we propose an approach to reduce the average number of samples (ASN) required by the Truncated Sequential Energy Detector (TSED) for Cognitive Radio. The reduction in overall ASN is achieved by forcing as many decisions as possible early on during the detection process with the tolerable error limits as the constraints. The sequential thresholds for each sample point are individually computed a priori using the probability distribution functions (PDF) of the test statistic values. The simulation results show that for comparable error performance, the proposed approach with individually computed thresholds reduces the ASN required by TSED to over 50% of the fixed sample size (FSS) required by conventional energy detector (ED) under high SNR.