The use of subvector quantization and discrete densities for fast GMM computation for speaker verification

Last year, we showed that the computation of a GMM-UBMbased speaker verification (SV) system may be sped up by 30 times by using a high-density discrete model (HDDM) on the NIST 2002 evaluation task. The speedup was obtained using a special case of the product-code vector quantization in which each dimension is scalar-quantized in the construction of the discrete model. However, the speedup resulted in a drop of an

• Type: Journal
• Year: 2010
• Where: INTERSPEECH
• Authors: Guoli Ye, Brian Mak
• Keywords: Discrete Model, GMM-UBMbased Speaker Verification, INTERSPEECH 2010, NIST 2002 Evaluation, Signal Processing