In this paper, we examine the problem of quality measurement for speaker verification using support vector machines (SVMs). An efficient Gaussian mixture models (GMMs) based quality estimation algorithm is proposed to potentially utilize speaker-specific broad acoustic-class characteristics. Some verification strategies are also considered in the test phase. We perform clustering-based vector pre-quantization to reduce the computational load and the redundancy in speech signal. Quality estimation is also integrated into test-length normalization. We then apply it to a text-independent speaker verification task using the NIST 2002 speaker recognition evaluation (SRE) database. Experimental results show that the proposed method can produce good classification accuracy. Keywords: Speaker verification, quality measure, vector pre-quantization, support vector machines, Gaussian mixture models.