曹洪林 李敬阳 王英利 孔江平

【摘 要】 本文针对目前国内外讨论比较热烈的声纹鉴定意见表述问题进行了评述。 首先介绍了实践中正在使用的听觉分析法、声谱比对分析法、声学分析法、听觉-声学分析法和说话人自动识别五种鉴定方法，指出了各种方法的优缺点；然后对现存的二元判决、可能性等级、似然比和英国立场声明四种鉴定意见表述形式进行了介绍和评析，通过分析发现，上述四种意见表述形式都存在一定的问题，实践中选择何种形式表述鉴定意见要综合考虑其科学性、逻辑性、现实性和可行性等多种价值选项；最后认为解决该问题的根本方法是各相关领域的专家应加强在鉴定方法上的合作性的基础研究。

【关键词】 声纹鉴定；鉴定意见；可能性等级；似然比；英国立场声明；贝叶斯原理；价值选择。

【中图分类号】 D915.13

【文献标识码】 A

【文章编号】 1674-1226（2013）05-0605-20

On Expert Opinion of Forensic Speaker Identification. Cao Honglin1,2 Li Jingyang3,4 Wang Yingli5 Kong Jiangping1 1 Department of Chinese Language and Literature, Peking University, Beijing, China. 100871; 2 Institute of Evi- dence Law and Forensic Science, China University of Political Science and Law, Beijing, China.100088; 3 Key Laboratory of Intelligent Speech Technology, Ministry of Public Security, Hefei, China. 230061; 4 Institute of Forensic Science, Ministry of Public Security, Beijing, China. 100038; 5 Center of Criminal Technology, Public Security Bureau of Guangdong Province, Guangzhou, China. 510050

【Abstract】 This paper is concerned with a critique of how to express the expert opinion (conclusion) on forensic speaker identification (comparison), about which considerable discussion has been made. First, five approaches that have been practiced currently, namely, auditory approach, spectrographic approach, acoustic approach, auditory-acoustic approach and automatic speaker recognition are reviewed, and the merits and demerits are pointed out respectively. Then four frameworks of expert opinion of forensic speaker identification, namely, binary decision, probability scale, likelihood ratio and UK Position Statement are presented and commented. The authors argue that all the four frameworks are problematic somehow when they are put into implementation, and thus we contend that scientificity, logicality, reality, feasibility and other value choices should be balanced before choosing any types of expert opinions. In conclusion, we believe that basic and cooperative research should be strengthened by experts with related backgrounds on improving the forensic speaker identification approach of analysis, and maybe this is the fundamental method for solving this problem.

【Key Words】 forensic speaker identification， expert opinion， probability scale， likelihood ratio， UK Position Statement， Bayes’ Theorem, value choice