Unknown radar emitter signal (RES) recognition is an important issue in modern electronic warfare because the enemy's RESs are usually uncertain in the battlefield. Although unsupervised classifiers are used generally in many domains, few literatures deal with applications of unsupervised classifiers to RES recognition. In this paper, three unsupervised classifiers including competitive learning neural network (CLNN), self-organizing feature map neural network (SOMNN) and support vector clustering (SVC) are used to recognize unknown RESs. 135 RESs with 7 intra-pulse modulations are used to test the performances of the three classifiers. Experimental results show that SVC is only slightly superior to CLNN and is greatly inferior to SOMNN.