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。     

Effects of language mismatch in automatic forensic voice comparison using deep learning embeddings

In forensic voice comparison, deep learning has become widely popular recently. It is mainly used to learn speaker representations, called embeddings or embedding vectors. Speaker embeddings are often trained using corpora mostly containing widely spoken languages. Thus, language dependency is an important factor in automatic forensic voice comparison, especially when the target language is linguistically very different from that the model is trained on. In the case of a low-resource language, developing a corpus for forensic purposes containing enough speakers to train deep learning models is costly. This study aims to investigate whether a model pre-trained on multilingual (mostly English) corpus can be used on a target low-resource language (here, Hungarian), not represented by the model. Often multiple samples are not available from the offender (unknown speaker). Samples are therefore compared pairwise with and without speaker enrollment for suspect (known) speakers. Two corpora are used that were developed especially for forensic purposes and a third that is meant for traditional speaker verification. Speaker embedding vectors are extracted by the x-vector and ECAPA-TDNN techniques. Speaker verification was evaluated in the likelihood-ratio framework. A comparison is made between the language combinations (modeling, LR calibration, and evaluation). The results were evaluated by Cllr_min and EER metrics. It was found that the model pre-trained on a different language but on a corpus with a significant number of speakers can be used on samples with language mismatch. Sample duration and speaking style also seem to affect the performance.     

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。     

A parameter study regarding the IBIS correlator

The electronic system IBIS has been used by numerous agencies worldwide as the standard tool to compare firearm markings on bullets and fired cartridges. There is a general interest among users concerning the likelihood with which the IBIS correlator may locate hits in its databases. Test results of the performance under different test conditions have been published in various papers. Experience has also been gained with the IBIS system from years of practical usage. All of these findings are difficult to compare with each other. No systematic presentation exists that actually shows the parameters upon which the success rate of the IBIS correlator depends. There has also been no mention of what values these parameters take on during each test. This paper first generally defines the success and error rates of the IBIS correlator. The parameters used will be discussed. Results of previously published tests will be re-examined based on this methodology. An illustrative form of presentation for the success rate of an electronic comparison system will also be suggested. It will be shown that the success rate of the IBIS correlator highly depends on the quality of the firearms-generated markings. It increases with the number of considered mark types, the number of available signatures per firearm, and the number of items inspected in the hit list. The success rate decreases with the database size. The paper will conclude with a series of practical recommendations for the setting up and successful operation of an electronic collection of ballistic evidence.     

Speech Watermarking: An Approach for the Forensic Analysis of Digital Telephonic Recordings*

Abstract: In this article, the authors discuss the problem of forensic authentication of digital audio recordings. Although forensic audio has been addressed in several articles, the existing approaches are focused on analog magnetic recordings, which are less prevalent because of the large amount of digital recorders available on the market (optical, solid state, hard disks, etc.). An approach based on digital signal processing that consists of spread spectrum techniques for speech watermarking is presented. This approach presents the advantage that the authentication is based on the signal itself rather than the recording format. Thus, it is valid for usual recording devices in police‐controlled telephone intercepts. In addition, our proposal allows for the introduction of relevant information such as the recording date and time and all the relevant data (this is not always possible with classical systems). Our experimental results reveal that the speech watermarking procedure does not interfere in a significant way with the posterior forensic speaker identification.     

The Classification of Inkjet Inks Using AccuTOF™ DART™ (Direct Analysis in Real Time) Mass Spectrometry—A Preliminary Study

A novel approach for the analysis of inkjet inks is being reported. A time‐of‐flight mass spectrometer, coupled with a Direct Analysis in Real Time (DART?) ion source (AccuTOF? DART?), was used to determine if inkjet inks from various manufacturers and models of printers could be reliably differentiated, characterized, and identified. A total of 217 ink standards were analyzed. As inkjet printing often involves the use of multiple colors (e.g., cyan, magenta, yellow, and black) to form an image or text, two different approaches to creating a library of standards and sampling methods were evaluated for implementation in a standard operating procedure. This research will show that a microscopic examination of the region of interest is requisite to identify what colors were utilized during the printing process, prior to comparing with known standards. Finally, blind testing was administered with 10 unknown samples to assess the validity and accuracy of the methodology.     

论声纹鉴定意见的表述形式

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.     

Forensic Analysis of Explosives Using Isotope Ratio Mass Spectrometry (IRMS)—Part 1: Instrument Validation of the DELTAplusXP IRMS for Bulk Nitrogen Isotope Ratio Measurements

Abstract: A significant amount of research has been conducted into the use of stable isotopes to assist in determining the origin of various materials. The research conducted in the forensic field shows the potential of isotope ratio mass spectrometry (IRMS) to provide a level of discrimination not achievable utilizing traditional forensic techniques. Despite the research there have been few, if any, publications addressing the validation and measurement uncertainty of the technique for forensic applications. This study, the first in a planned series, presents validation data for the measurement of bulk nitrogen isotope ratios in ammonium nitrate (AN) using the DELTA^plusXP (Thermo Finnigan) IRMS instrument equipped with a ConFlo III interface and FlashEA? 1112 elemental analyzer (EA). Appropriate laboratory standards, analytical methods and correction calculations were developed and evaluated. A validation protocol was developed in line with the guidelines provided by the National Association of Testing Authorities, Australia (NATA). Performance characteristics including: accuracy, precision/repeatability, reproducibility/ruggedness, robustness, linear range, and measurement uncertainty were evaluated for the measurement of nitrogen isotope ratios in AN. AN (99.5%) and ammonium thiocyanate (99.99+%) were determined to be the most suitable laboratory standards and were calibrated against international standards (certified reference materials). All performance characteristics were within an acceptable range when potential uncertainties, including the manufacturer’s uncertainty of the technique and standards, were taken into account. The experiments described in this article could be used as a model for validation of other instruments for similar purposes. Later studies in this series will address the more general issue of demonstrating that the IRMS technique is scientifically sound and fit‐for‐purpose in the forensic explosives analysis field.     

从话者识别角度认识山西汾城方言的紧喉音节

目的明确山西汾城方言紧喉音节与对应普通话音节在司法话者识别时的可比性及该紧喉音的声学特性。方法利用VS-99语音工作站对山西汾城方言紧喉音节及对应普通话音节的韵律特性、音质特性等声学特性进行定性与定量分析。结果得出了汾城方言紧喉音的发声态主要为嘎裂声,汾城方言紧喉音对声母没有明显影响,紧喉音韵母与普通话韵母在共振峰频率上的一致性要远高于共振峰相对强度上的一致性等认识。结论为在汾城方言与普通话之间进行司法话者识别以及田野调查提供了依据。     

Assessing Cognitive Bias in Forensic Decisions: A Review and Outlook

In recent years, a number of studies have demonstrated that forensic examiners can be biased by task-irrelevant contextual information. However, concerns relating to methodological flaws and ecological validity attenuate how much the current body of knowledge can be applied to real-life operational settings. The current review takes a narrative approach to synthesizing the literature across forensic science. Further, the review considers three main issues: (i) primary research on contextual bias within forensic science; (ii) methodological criticisms of this research; (iii) an alternative perspective that task-irrelevant contextual information does not always lead to error. One suggestion for future research is outlined, which is that studies on contextual bias in forensic decisions should be conducted in collaboration between forensic scientists and cognitive psychologists. Only then can rigorous and ecological valid experiments be created that will be able to assess how task-irrelevant contextual information influences forensic analysis and judgments in operationally valid settings.     

Advanced Framework for Digital Forensic Technologies and Procedures*

Abstract: Recent trends in global networks are leading toward service‐oriented architectures and sensor networks. On one hand of the spectrum, this means deployment of services from numerous providers to form new service composites, and on the other hand this means emergence of Internet of things. Both these kinds belong to a plethora of realms and can be deployed in many ways, which will pose serious problems in cases of abuse. Consequently, both trends increase the need for new approaches to digital forensics that would furnish admissible evidence for litigation. Because technology alone is clearly not sufficient, it has to be adequately supported by appropriate investigative procedures, which have yet become a subject of an international consensus. This paper therefore provides appropriate a holistic framework to foster an internationally agreed upon approach in digital forensics along with necessary improvements. It is based on a top‐down approach, starting with legal, continuing with organizational, and ending with technical issues. More precisely, the paper presents a new architectural technological solution that addresses the core forensic principles at its roots. It deploys so‐called leveled message authentication codes and digital signatures to provide data integrity in a way that significantly eases forensic investigations into attacked systems in their operational state. Further, using a top‐down approach a conceptual framework for forensics readiness is given, which provides levels of abstraction and procedural guides embellished with a process model that allow investigators perform routine investigations, without becoming overwhelmed by low‐level details. As low‐level details should not be left out, the framework is further evaluated to include these details to allow organizations to configure their systems for proactive collection and preservation of potential digital evidence in a structured manner. The main reason behind this approach is to stimulate efforts on an internationally agreed “template legislation,” similarly to model law in the area of electronic commerce, which would enable harmonized national implementations in the area of digital forensics.     

APPLYING CASE STUDY METHODOLOGY TO CHILD CUSTODY EVALUATIONS

The child custody evaluation (CCE) can play a critical role in the resolution of custody disputes in the Family Court. There have been a number of significant methodological advances made in this field by various researchers over the past 20 years. It is timely that a scientifically‐based, integrative, methodological framework be developed within which existing methods might be situated. In this paper, case study methodology is proposed as an appropriate methodological framework for CCEs. The application of this methodology to CCEs is explicated with particular attention being paid to the methodological tasks of data collection and data interpretation. An orienting model is proposed to guide the collection of data, and strategies are described for applying population level research findings to individual cases in the form of risk and resilience models. Finally, coherence, analogy, and making methodology explicit are proposed as evaluation criteria for the CCE process and reports, and suggestions are made about outcome research.     

Voice stress evaluators and lie detection

Proponents of vocal stress analysis systems argue that they are able to detect spoken deception by analysis of "stress" in the voice signal. Presumably, they do so by examining traces made by laryngeal microtremors which, they claim exist in the voice, are associated with stress, and ultimately are associated with lying. However, most research that seeks to identify the relationships between microtremors and laryngeal function has produced negative results, and data on the ability of voice analyzers to detect stress from speech--or to identify spoken deception--have been negative or "mixed" in nature. Since perspectives based on available results leave a number of questions unanswered, a series of experiments has been undertaken. The first was focused on the basic acoustic/temporal correlates of stress in voice (the subject of an earlier report), the second on examination of stress by commercial voice analyzers, and the third on the detection of relatively high-risk lies by this same type of voice analysis procedure. It was found that correct stress/nonstress identifications occurred only at chance levels; the lie/nonlie identification scores were quite similar with professional "examiners" performing at about the same level of accuracy as other auditors. The following review is divided into two parts: a history of the controversy and a presentation of the two cited experiments.     

Cranial Nonmetric Variation and Estimating Ancestry*

Abstract: Historically, when predicting the ancestry of human skeletal remains, forensic anthropologists have not fully considered the variation within human populations, but instead have relied on a typological, experience‐based approach. Unfortunately, reliance on observer experience has produced a method that is as much an art as it is a science. This research focuses on the frequency distribution and inter‐trait correlations of 11 common morphoscopic traits to demonstrate that the experience‐based approach to ancestry prediction is indeed an art that is unscientific, because it is unreplicable, unreliable, and invalid. Ten of 11 traits examined had frequency distributions with significant differences (p < 0.001) between groups, but the range in variation of these traits far exceeds previous assumptions. Such within group variation clearly demonstrates that extreme trait expressions are not reliable for estimating ancestry through visual observation alone, but instead that these traits should be analyzed within a statistical framework.     

Introducing Human Mandible Identification [(hu)MANid]: A Free,Web‐Based GUI to Classify Human Mandibles

Statistical programs have revolutionized the way in which forensic anthropologists conduct casework by allowing practitioners to use computationally complex analytics at the click of a button. Importantly, the products of these statistical programs are reproducible and contain measures of error or uncertainty, thereby strengthening conclusions. This paper is an introduction to (hu)MANid, a free, web‐based application that uses linear and mixture discriminant function analyses to classify human mandibles into one of many worldwide and/or periodic reference groups. The mechanics, development, and use of the application will be discussed. Further, the program was tested against other software to compare model performances and classifications. Total correct classifications among the test cases and programs were identical. Ten mandibles were tested using both statistical procedures. Mixture discriminant analysis improved classification by an average of 9.3% and correctly identified three more mandibles than LDA. Therefore, we believe (hu)MANid will be an asset to the anthropological community.     

Stress and Deception in Speech: Evaluating Layered Voice Analysis*

Abstract: This study was designed to evaluate commonly used voice stress analyzers—in this case the layered voice analysis (LVA) system. The research protocol involved the use of a speech database containing materials recorded while highly controlled deception and stress levels were systematically varied. Subjects were 24 each males/females (age range 18–63 years) drawn from a diverse population. All held strong views about some issue; they were required to make intense contradictory statements while believing that they would be heard/seen by peers. The LVA system was then evaluated by means of a double blind study using two types of examiners: a pair of scientists trained and certified by the manufacturer in the proper use of the system and two highly experienced LVA instructors provided by this same firm. The results showed that the “true positive” (or hit) rates for all examiners averaged near chance (42–56%) for all conditions, types of materials (e.g., stress vs. unstressed, truth vs. deception), and examiners (scientists vs. manufacturers). Most importantly, the false positive rate was very high, ranging from 40% to 65%. Sensitivity statistics confirmed that the LVA system operated at about chance levels in the detection of truth, deception, and the presence of high and low vocal stress states.     

A GIS‐based Quantitative Approach for the Search of Clandestine Graves,Italy

Previous research on the RAG color‐coded prioritization systems for the discovery of clandestine graves has not considered all the factors influencing the burial site choice within a GIS project. The goal of this technical note was to discuss a GIS‐based quantitative approach for the search of clandestine graves. The method is based on cross‐referenced RAG maps with cumulative suitability factors to host a burial, leading to the editing of different search scenarios for ground searches showing high‐(Red), medium‐(Amber), and low‐(Green) priority areas. The application of this procedure allowed several outcomes to be determined: If the concealment occurs at night, then the “search scenario without the visibility” will be the most effective one; if the concealment occurs in daylight, then the “search scenario with the DSM‐based visibility” will be most appropriate; the different search scenarios may be cross‐referenced with offender's confessions and eyewitnesses’ testimonies to verify the veracity of their statements.     

中外法庭科学标准化研究现状比较与发展趋势

近年来,我国法庭科学发展迅猛,开始受到广泛关注和重视。经过多年的研究和推动,标准化正逐步深入到法庭科学的日常实践中,通过制定和实施标准规范,确保法庭科学结果质量,更好发挥科学技术在执法司法中的重要作用。标准化研究是法庭科学标准化工作的重要环节,也是推动法庭科学标准化发展的基础和动力。国内外法庭科学研究者都高度重视法庭科学标准化研究,许多新理念、新思想、新方法不断涌现。本文从标准定位、标准化对象、标准颗粒度、标准体系、质量控制、国际标准化六个方面对当前中外法庭科学标准化研究的现状进行了深入全面的归纳和比较,指出了国内外在法庭科学标准化研究方面的相似和差异,并对法庭科学标准化研究的未来发展趋势进行了分析,以求为推动我国法庭科学领域标准化工作高质量发展提供借鉴和参考。     

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。     

Reconstruction of Deformed Bite Marks Using Affine Transformations

Abstract: Bite marks inflicted on animate and inanimate objects will undergo a certain degree of deformation. This fact remains one of the biggest stumbling blocks when analyzing evidence for court presentation. It has been demonstrated that the reliability of pattern association analysis will not be affected by minimal degrees of warping, shrinkage, and distortion. In this study, affine transformations were applied to bite marks to establish if minimal distortions would affect the mathematically determined relationships of the defined features. In a real case study, it was then tested whether the distorted bite mark matched the dentition of the perpetrator by applying an affine transformation. This was confirmed to be the case. Affine transformations will thus not affect the relationships of the individual features found in bite marks. The numerical calculations validate the reliability of pattern association analysis in the presence of minimal amounts of warping, shrinkage, and distortion.