To cope with the issue of the idle state detection which is difficult in motor imagery based brain-computer interface, the paper proposes a method that approximate entropy and common spatial pattern(CSP) are combined. On the condition of two-class classification, different kings of features are extracted through approximate entropy in time complexity and CSP in spatial pattern. Then these features are used to make different classifiers which are combined by vote-based classification method to improve the accuracy of judging idle state. By way of this method, the final experimental results of BCI competition shows the true positive rate(TPR) of intentional motor imagery is higher than the threshold method. The result of data processing indicates the effectiveness of the proposed method.