司法语音及声学研究.doc

【法宝引证码】 CLI.A.1239700? 原文链接： HYPERLINK /qikan/2ddd24f832efade1115a109edc99ab8ebdfb.html \t _blank /qikan/2ddd24f832efade1115a109edc99ab8ebdfb.html 2017年司法语音及声学研究 期刊名称： HYPERLINK /journal/articlebyclass?ClassCodeKey=172, \t _blank 《刑事技术》 期刊年份： HYPERLINK /journal/findarticle?FilterItems.JournalYear=2018 \t _blank 2018 摘要： 司法语音及声学在2017年发展迅速，本文对其中的代表性成果进行梳理。在语音同一认定方面，VPA成为听觉分析的热点，一些机构对鉴定意见表述分类更加细致，自动识别技术不断发展，其与专家鉴定的关系各方仍有不同意见，部分机构与学者开始推动检验过程向更加透明化的方向发展。在语音人身分析方面，除了传统的对性别、年龄、地域、体形等属性的推断，语音情感分析也成为研究热点，无论是专家分析还是自动分析都取得了进步。录音的真实性检验中，除了传统检验流程的深入研究，电网频率仍是关注热点。降噪及语音增强开始使用深度神经网络技术并取得了一定成果。 英文摘要： Forensic phonetics and acoustics have made fast development in 2017. This paper gives a short introduction to their state-of-the-art techniques. In forensic speaker identification (FSI), many scholars have made efforts in standardizing the procedures of auditory analysis. These procedures are being used in several forensic speaker comparison laboratories around the world. Besides, Vocal profile analysis (VPA) is the focus of the techniques and several scholars have devised different versions of VPA for forensic purpose, and its utility even extends to the optimization of automatic speaker comparison systems. Phonetic-acoustic analysis has been broadened in different directions, including filled pauses, diphthongs and voice profiling. The value of diverse vocal characteristics is measured in different ways; concepts like relevant population and homogeneity measurement are initiated to quantize accurately. Evaluation system is divided into more scales in some institutions for various reasons like avoiding cliff-edge effect and keeping accordance with other forensic counterparts. Automatic speaker recognition moves forward although there are controversies over its relationship with experts and its role in FSI. Whether conclusions from these machines should be accepted in court as evidence is argued around the world and yielding different attitudes. Some institutions and profession