Aiming at the spectrum sensing demand under low SNR environment of cognitive radio (CR) users, a cooperative spectrum sensing algorithm based on truncated sequential probability radio test (SPRT) is proposed. Each cooperative radio user counts the number of sampling points, while its energy value is greater than the preset threshold in every field by sectionally processing the signal sequences they received, and then upload them to the fusion center as local test statistics, which reduces the overhead of control channel effectively. By introducing De Moivre-Laplace theorem and central limit theorem, the local test statistics of intrasegment approximate Gaussian distribution, which greatly simplifies the theoretical deduction and calculative process of likehood ratio function subsequently. Performance analysis and simulation results show that under the same conditions of detection performance, the proposed algorithm reduces the spectrum sensing time largely compared with the conditional energy detection.