CAJ | 학술논문

为减弱注册语音与测试语音时长不一致对说话人识别性能的负面影响,提出一个概率修正PLDA建模方法.根据语音时长自适应改变传统PLDA模型中i-vector的概率分布函数,提高PLDA对每个说话人每段语音的时长表征能力,以增强说话人类别的区分度.为验证基于概率修正PLDA模型的有效性,进行了NIST SRE10 core-core测试集在3种不同时长的评测实验,以及NIST 2014 i-vector machine learning challenge测试任务.结果表明,相较于传统的PLDA训练模型,通过语音时长的约束提高了说话人识别性能.
위감약주책어음여측시어음시장불일치대설화인식별성능적부면영향,제출일개개솔수정PLDA건모방법.근거어음시장자괄응개변전통PLDA모형중i-vector적개솔분포함수,제고PLDA대매개설화인매단어음적시장표정능력,이증강설화인유별적구분도.위험증기우개솔수정PLDA모형적유효성,진행료NIST SRE10 core-core측시집재3충불동시장적평측실험,이급NIST 2014 i-vector machine learning challenge측시임무.결과표명,상교우전통적PLDA훈련모형,통과어음시장적약속제고료설화인식별성능.
To reduce the negative impact on the performance of speaker recognition systems due to the duration mis-match between enrollment utterance and test utterance,a modified-prior PLDA method is proposed.The probability distribution function of i-vector was modified by incorporating the covariance matrix with duration of each utterance of each speaker during the PLDA training,which further improved the discriminant capability of speaker classifica-tion.To evaluate the robustness of the proposed modified-prior PLDA method,extensive experiments were per-formed on NIST SRE10 core-core task(female part)in duration mismatch conditions and NIST 2014 i-vector machine learning challenge.Experimental results demonstrated that the duration-based modified-prior PLDA method achieved better performance compared with the traditional PLDA.