伪装语音的自动话者识别研究

本文以语音自动话者识别系统为平台,对实际案件中出现频率较高的10种伪装语音类型进行自动话者识别测试。通过对20位发音人的正常语音和10种类型伪装语音的话者辨认和话者确认测试,分析不同语音伪装类型对自动话者识别的影响。该结果对于深入认识伪装语音的特性及其话者识别研究具有重要意义,也为自动话者识别技术的改进提供了参考。     

语音身份识别及其在犯罪侦查中的运用

语音身份识别是通过说话人的语音分析其性别、年龄、职业、文化程度等,从而推断说话人身份的一种人身识别技术。语音身份识别一般需通过听辨和视图检测,有的还需聘请专家会检。语音身份识别在刑事犯罪中可以确定侦查方向、缩小犯罪嫌疑人的范围,因此,在绑架、敲诈勒索、诈骗等案件中有较高运用价值。     

不同通信系统下的说话人识别探究

目的研究不同通信系统条件下,利用语音进行说话人识别的基本假设＂个体之间的语音差异大于个体自身的语音变异＂是否成立。方法选择宽带语图差异较大的不同通信系统下语音进行比对及差异的统计。结果发现说话人识别的基本假设在不同通信系统条件下成立。结论得出了不同通信系统条件下进行说话人识别的方法与判阈。     

国内语言识别领域专利技术发展趋势

前言语音技术在计算机领域中的关键技术是语音识别技术和语音合成技术,其中语音识别技术是2000年至2010年间信息技术领域十大重要的科技发展技术之一。语音识别技术(ASR:AutomaticSpeechRecognition),是指将人说话的语音信号转换为可被计算机程序所识别的文字信息,从而识别说话     

辨证认识话者自动识别系统

为了警示近年来我国司法话者识别领域中出现的一些崇外、盲目追求快速与省事的苗头,结合话者自动识别系统的研究、应用状况,从语音的共性与个性、话者识别结果的相对性与绝对性出发,通过分析比对话者自动识别与语音识别所用的特征参数及实现过程,辨证分析了制约话者自动识别系统准确率的根本原因。指出了话者自动识别系统尚无法达到人们对其的期望,以及适合于司法诉讼领域的话者自动识别系统的发展方向。     

噪声环境下的说话人识别研究

论文研究了噪声环境下的基于GMM模型和MFCC参数的说话人识别方法.白噪声环境下,说话人识别的正确率大大降低.论文研究了GMM模型在说话人识别中的应用,并采用MFCC参数作为语音的特征参数进行噪声环境下的识别研究.实验结果表明,在0dB到15dB的噪声环境下,使用GMM模型进行说话人识别时,采用MFCC参数比采用LPCC特征参数能够获得更高的识别率.在10dB的低信噪比环境下,使用MFCC参数的说话人识别,仍然可以获得75.6％的识别率,可以有效进行识别.     

法庭语音学

本文对法庭语音学进行了研究综述,主要介绍了该学科的核心内容:说话人鉴定。在实际办案中,当未找到嫌疑人,只有犯罪分子的检材语音时,可以使用说话人画像/说话人分类技术。若没有犯罪分子的录音证据时,可以让受害人和证人进行说话人的听觉辨认。具体的辨认形式有两种:对熟人辨认和对陌生人辨认,在对陌生人辨认时可以采用语音辨认的方法进行。当检材语音和样本语音都齐备的时候,法庭语音分析专家就可以对二者进行比对检验了。目前语音比对分析涉及到的问题和领域有:基于贝叶斯方法的法庭推理和似然比计算、共振峰频率的测量应用、非解析感知与样例理论、法庭说话人自动识别以及不同方法的综合应用等。     

法庭语音学

本文对法庭语音学进行了研究综述,主要介绍了该学科的核心内容：说话人鉴定。在实际办案中,当未找到嫌疑人,只有犯罪分子的检材语音时,可以使用说话人画像/说话人分类技术。若没有犯罪分子的录音证据时,可以让受害人和证人进行说话人的听觉辨认。具体的辨认形式有两种：对熟人辨认和对陌生人辨认,在对陌生人辨认时可以采用语音辨认的方法进行。当检材语音和样本语音都齐备的时候,法庭语音分析专家就可以对二者进行比对检验了。目前语音比对分析涉及到的问题和领域有：基于贝叶斯方法的法庭推理和似然比计算、共振峰频率的测量应用、非解析感知与样例理论、法庭说话人自动识别以及不同方法的综合应用等。     

利用线性预测谱LPC对元音区别力量化分析的研究

本研究利用LPC对10位发音人的正常单元音相同声调（a、o、e、i、u、ü）进行识别测试,进一步探讨LPC提取出来的各个参数对于区分说话人的作用。提取元音最平稳的一段进行长时LPC分析,分析不同人的语声声纹差异性和同一人语声声纹的总体特殊性。结果显示,每个话者的自身差异都明显小于话者之间的差异。因此利用LPC提取元音的共振峰振幅与相邻弱谷幅值之差对比分析对话者具有有效区别力。     

似然比在法庭说话人确认中的应用

本文把成功应用于DNA检验的证据评估方法（似然比）应用于法庭语音证据评估之中,提取语音的LPC作为识别特征,并使用45人电话对话录音中元音/a/作为样本进行了测试。结果表明该方法不仅能正确识别说话人,而且能根据当前嫌疑人样本和问题语音样本的差异,量化该语音样本作为证据的力度,为法庭提供科学合理的证据评估结果和科学解释。同时,自动特征提取的引入比起人工提取共振峰特征,提高了工作效率,识别系统性能也获得大幅提升。     

An esoteric technique useful in the identification of unidentified remains from the examination of faded,illegible hospital identification wristbands

A technique routinely used in the examination of questioned documents has been found to be of assistance when employed in the examination of faded and/or partially legible hospital identification wristbands found with unidentified remains (UIDs). A non-destructive test used predominately by forensic document examiners in the analysis of writing inks, handwritten alterations, and obliterations has proven useful throughout the years when confronted with this unusual type of documentary evidence. This discussion paper was prompted by the Tri-State Crematory disaster, Walker County, Georgia, from a request by investigators as to whether or not any information could be obtained from the examination of faded hospital identification wristbands where no information was readily discernable. Subsequent analysis by non-destructive infrared inspection, a standard technique used in the examination of questioned documents, proved useful in assisting with the identification of unidentified skeletal remains.     

Do Prophylactics Prevent Inflation? Post-identification Feedback and the Effectiveness of Procedures to Protect Against Confidence-inflation in Earwitnesses

After viewing or hearing a recorded simulated crime, participants were asked to identify the offender’s voice from a target-absent audio lineup. After making their voice identification, some participants were either given confirming feedback or no feedback. The feedback manipulation in experiment 1 led to higher ratings of participants’ identification certainty, as well as higher ratings on retrospective confidence reports, in both the immediate and delay groups. Earwitnesses who were asked about their identification certainty prior to the feedback manipulation (experiment 2) did not demonstrate the typical confidence-inflation associated with confirming feedback if they were questioned about the witnessing experience immediately; however, the effects returned after a week-long retention interval. The implications for the differential forgetting and internal-cues hypotheses are discussed.     

手机通话语音的实验研究

当前,手机通话语音已成为司法语音鉴定中最为常见的一种语音形式。本研究从手机通信系统的信道特点出发,分析手机通话语音的声谱特点和共振峰频率变化等情况;同时还比较了不同通话网络、不同通话方式及不同手机的通话语音特点。实验发现,手机通话语音与直接录音语音有明显的变化,主要表现在高低频信息的带宽滤波效应、高低频共振峰的漂移、语音质量、音色、韵律特征等方面;还发现,不同手机通话条件下的语音变化程度不同。最后,讨论了手机通话语音变化对说话人鉴定的影响及鉴定中的注意事项。     

Identification of human figure drawings through questioned document examination techniques

Two cases were submitted for questioned document examination involving human figure drawings. As there was little information on ‘artist’ identification in the questioned document examination literature, a sample of 107 individuals was selected to participate in a study. The respondents were requested to draw two cartoons, one naturally and one disguised, in order to determine if standard questioned document examination methods could be employed to identify the ‘artists’. The findings indicated that ‘artists’ of human figure drawings could be identified using standard handwriting identification techniques.     

再看摹仿签名笔迹鉴定

由于签名笔迹字数少、可摹仿性强的特点,长期以来,摹仿签名笔迹鉴定始终是笔迹鉴定的难点。笔者从不同类型的摹仿方法出发,讨论不同方法下摹仿签名笔迹的特点,总结其特征规律。同时要重视笔迹鉴定过程中了解相关案情,为识别是否存在摹仿现象提供线索。在此基础之上,对比摹仿签名笔迹特征的变化规律,并从摹仿签名笔迹特征反映构成方面系统性地进行鉴定,鉴别是否存在摹仿签名的情况。通过撰写此文,以期对摹仿签名笔迹鉴定实务具有借鉴意义。     

Inaccuracy of height information on driver's licenses.

Height is an important variable in identification. In cases involving decreased individuals, height is measured directly on the corpse or estimated from the skeleton. This postmortem measurement or estimation is then compared with antemortem records, usually the driver's license. The accuracy of the license information, however, has been questioned. To assess the accuracy of driver's license information, volunteers' statures were measured, and then these figures were compared with those printed on the subjects' licenses. Even in our comparatively young, well-educated sample, the license height was significantly greater than the measured height. Some inaccuracies may be caused by failure to update license information when new licenses are issued, but some inaccuracies may be from personal deception. The implications of these results for forensic anthropology cases are discussed.     

Long-term working memory in voice identification

Abstract

The present study investigated the role of long-term working memory in voice identification. For this purpose, a mock theft design was used and three groups: voice identification experts, totally blind people and a control group of sighted listeners without any special training or experience in voice identification, were compared for their voice identification efficiency.

Results revealed that the professional voice identification experts were more accurate than the other groups. Blind listeners were as effective as the sighted in voice identification. The results confirmed predictions which were based on the long-term working memory model. Implications of these results for the practice of voice identification are discussed.