Abstract (
)
Download PDF
HTML
(
)
Knowledge map
Save
To enhance the self-learning ability of models in acoustic scene classification tasks and reduce the cost of sample annotation, an improved semi-supervised model, ASC-FixMatch, is proposed in this study. The ASC-FixMatch model is capable of fully utilizing the limited labeled data and the abundant unlabeled data. Building upon the fundamental architecture of convolutional recurrent neural networks, the model has refined the consistency regularization strategy of FixMatch. It compels the learning of invariant features that remain stable after severe transformations by expanding the data perturbation space at the deep feature level, thereby enhancing the model's generalization ability to new data. During experiments, three different weighting strategies for the loss function of unlabeled data were adopted. The performance of the ASC-FixMatch model was compared with other classical semi-supervised models. The results indicated that, on the TAU Urban Acoustic Scenes 2019 development data set, the ASC-FixMatch model with the best loss weighting strategy achieves an accuracy improvement of 16.70%, 4.71%, and 6.65% compared to semi-supervised models based on pseudo-labeling, mean-teacher methods, and interpolation consistency training algorithms, respectively. The proposed model demonstrates a certain superiority in performance in the task of semi-supervised acoustic scene classification.
Abstract (
)
Download PDF
HTML
(
)
Knowledge map
Save
Electronic data forensics plays a crucial role in modern criminal investigations. However, with the continuous enhancement of smartphone security mechanisms, user privacy protection features－such as privacy spaces and system clones－have posed significant challenges to forensic work, particularly when privacy passwords are unavailable. To address this challenge, this study combines the UFED tool with manual analysis methods to conduct a comprehensive investigation into the data storage mechanisms of privacy spaces and system clones in four smartphone brands: OPPO, Redmi, Vivo, and Honor. The experimental results demonstrate that data in the privacy space and system clone of some devices are stored in an unencrypted format within the file system. When the file systems were analyzed using three domestic forensic tools, data from the privacy space and system clone of some devices were not successfully parsed, revealing the data protection mechanisms in smartphones from different brands, as well as the challenges faced by domestic forensic tools. This study provides new insights into improving domestic smartphone forensic methods and offers new approaches to address the current challenges in privacy protection data extraction, with significant practical application value.
Abstract (
)
Download PDF
HTML
(
)
Knowledge map
Save
The origins and divergence of Sino-Tibetan populations are pivotal subjects of study in linguistics, anthropology, and population genetics. While most previous research have focused on the Han and/or Tibetan populations, comprehensive studies on other Sino-Tibetan populations, such as the Qiang and Yi, are still insufficient. This study collected Y-SNP haplogroup and Y-STR haplotype data from 1151 unrelated male individuals across 12 Sino-Tibetan populations. Using methods such as principal component analysis (PCA), phylogenetic tree construction, and haplotype network analysis, this study systematically reveal the paternal genetic structure and differentiation among Sino-Tibetan populations. Analysis of 90 Y-SNP haplogroups reveals that haplogroups O-M175, D1-M174, C-M130, and N-M231 account for over 85.37% of the paternal haplogroups in the studied populations. Sub-haplogroups of each major haplogroup exhibited distinct distribution patterns across different regions and populations. PCA results indicate a strong correlation between genetic clustering, language family, and geographical distribution. Phylogenetic tree and haplotype network analyses, using 16 Y-STR markers, further support the correlation between genetic differentiation and geographical distribution, clarifying the genetic relationships among Sino-Tibetan groups. Through the combined analysis of Y-SNP and Y-STR markers, this study deeply explores the distribution, genetic substructure, and internal genetic relationships of Y-chromosome haplogroups in Sino-Tibetan populations. It provides new theoretical insights into population genetics and offers data support for Y-chromosome research applications.
Abstract (
)
Download PDF
HTML
(
)
Knowledge map
Save
This study applies deep learning technology to the CT image reading and diagnosis of nasal area fractures in forensic clinical identification, with the aim of improving the accuracy of fracture localization and differentiation between fresh and old fractures, reducing the rates of missed diagnosis and misdiagnosis, and achieving intelligent recognition of nasal fractures. A total of 600 cases of confirmed fresh fractures, old fractures, and no fractures in the nasal area were collected. 9 294 frames of transverse and coronal CT images were captured and manually annotated for the overall bone structure and fracture site of the nasal area. A Faster R-CNN neural network, VGG16 backbone feature extractor, and CBAM module were used to establish an intelligent system model. The CT transverse images of the nasal area as a whole, the CT transverse images of the fracture site, and the CT coronal images of the fracture site were trained and tested separately. The results showed that the average recognition accuracy of the model in all three groups exceeded 90%, achieving accurate localization of fracture sites. This study confirms that the deep learning-based intelligent identification system model for nasal fractures can effectively assist forensic appraisers in CT image diagnosis of nasal fractures. It contributes to reduce the missed diagnosis and misdiagnosis of such injuries, and further improve the accuracy of forensic clinical practice.
Abstract (
)
Download PDF
HTML
(
)
Knowledge map
Save
In this study, multiple analytical techniques were employed to confirm the structure of a seized synthetic cannabinoid precursor and establish its characterization data, including gas chromatography-mass spectrometry (GC-MS), ultra high performance liquid chromatography quadrupole-orbitrap high resolution mass spectrometry(UPLC-Q Exactive HRMS), nuclear magnetic resonance spectroscopy (NMR) and infrared spectroscopy(IR). The compound was identified as 1-(4-fluorobenzyl) indazole-3-carboxylic acid, a precursor to the synthetic cannabinoid AB-FUBINACA. The method is scientific and efficient and the result is accurate and reliable. The integrated use of multiple techniques overcomes challenges such as the absence of reference substances and the lagging of spectrum database. This approach provides a new way to confirm the structure of precursor chemicals used in drug synthesis.
Abstract (
)
Download PDF
HTML
(
)
Knowledge map
Save
A method for rapidly detecting florfenicol in chicken breast was established based on magnetically dispersed solid-phase extraction-colloidal gold immunochromatography. This method integrates the efficient separation capabilities of magnetically dispersed solid-phase extraction and the rapid and sensitive characteristics of colloidal gold immunochromatography, enabling the rapid and accurate detection of florfenicol in chicken breast. By optimizing the pretreatment method, the extraction and colloidal gold immunochromatographic detection conditions were determined, and the established rapid detection method was applied to the chicken breast matrix. The experimental results demonstrated that the method exhibits excellent accuracy, high sensitivity, and good specificity.
Abstract (
)
Download PDF
HTML
(
)
Knowledge map
Save
With the rapid development of telecommunications technology, new forms of crimes such as illegal cyberattacks and telecom fraud keep emerging, which greatly raises the difficulty of investigations. The importance of cross-regional and cross-departmental collaborative case-solving has become increasingly prominent. To address the limitations of existing collaborative platforms in electronic evidence examination, including difficulties in data processing, restricted collaborative functionality, and insufficient automated analysis capabilities, this study innovatively designs and develops a new collaborative platform for case-related evidence materials. The platform employs a multi-language hybrid architecture integrating core modules such as electronic data transmission, BCP packet parsing, data cleaning, and batch extraction. Key innovations include modular design, one-click deployment in standalone environments, keyword extraction algorithms, and word cloud visualization. Application results verified that the platform significantly enhances case investigation efficiency and data analysis capabilities, providing real-time, mobile, and convenient case evidence disposal.
Abstract (
)
Download PDF
HTML
(
)
Knowledge map
Save
In motorcycle-car collision accidents, determining collision speed is critical for post-accident responsibility assessment. Traditional speed estimation methods mainly rely on skid marks and vehicle deformation analysis, which face limitations in accuracy and practicality. To address the shortcomings, this study proposes a dynamic calculation method for motorcycle speed in lateral collisions based on Event Data Recorder (EDR) data. A coordinate system centered on the car was established via the 3-2-1 method. Momentum conservation laws and impulse-momentum theorem were applied to construct a speed calculation model along the motorcycle’s travel direction. Key parameters, including pre-collision and post-collision velocities as well as the interaction duration, were extracted from EDR data to solve the model. The results demonstrate that this method reduces reliance on accident scene-based physical evidence.
Abstract (
)
Download PDF
HTML
(
)
Knowledge map
Save
This study investigates a novel method for developing latent fingerprints by combining an ultrasonic atomization device with an aggregation-induced emission (AIE) material, TPA-1OH. Experiments were conducted to determine the operational parameters of the ultrasonic atomization device, optimize the formulation ratio of the developer TPA-1OH, and evaluate its potential impact on subsequent DNA analysis. The results demonstrate the feasibility of the proposed method for latent fingerprint development using the ultrasonic atomization device integrated with TPA-1OH. Following treatment with TPA-1OH, latent fingerprints exhibited significantly enhanced visualization under a 445 nm laser with a yellow filter, outperforming untreated controls in both clarity and contrast. Additionally, no significant DNA degradation was detected in the developer-treated specimens, indicating compatibility with downstream forensic DNA workflows.
Abstract (
)
Download PDF
HTML
(
)
Knowledge map
Save
The research progress of artificial intelligence (AI) in forensic pathology was systematically reviewed, which focuses on the key technical challenges frequently encountered in forensic practice, including histopathological diagnosis, cause of death analysis, postmortem interval estimation, injury mechanism analysis, and bloodstain pattern analysis. By sorting out the intersections of forensic pathology examination and AI, this study aims to promote the application of relevant research findings in forensic practice and empowers the formation of new-quality combat effectiveness in public security, as well as the improvement of the intelligence level of forensic work.
Abstract (
)
Download PDF
HTML
(
)
Knowledge map
Save
Tracking code inspection technology, a novel forensic science technique, enables rapid, efficient, and non-destructive tracing of color laser-printed or copied documents to their source devices while detecting alterations such as text insertions, modifications, or page replacements. Over recent years, research on tracking codes has advanced from early investigations, such as visualization methods and morphological features of tracking dots, to current studies on their encoding principles, matrix patterns, automated extraction, and practical applications, yielding significant progress. This paper systematically reviews five key areas, namely visualization techniques for printer tracking codes, morphological and dimensional characteristics of tracking dots, periodic arrangement patterns of tracking dot matrices, automated extraction of tracking code periodicity, and decoding methodologies and applications. It concludes that informatization and database development represent critical directions for future advancements in tracking code research and forensic document examination.
Abstract (
)
Download PDF
HTML
(
)
Knowledge map
Save
The proliferation of illicit drugs poses a significant threat to public health and societal security. In this context, Raman spectroscopy has emerged as a promising tool for drug detection, offering non-invasive, rapid, and accurate analysis. This review explores the diverse applications and advancements of Raman spectroscopy in drug detection, including qualitative and quantitative analysis of unknown substances to determine drug content, identification of mixtures and adulterants in seized materials, on-site identification and rapid detection at crime scenes, detection of microquantities of drugs; and drug detection both in vivo and in vitro for suspected drug abusers. Additionally, the review addresses the application of Raman spectroscopy in monitoring drug manufacturing processes and the surveillance of illicit drugs precursor chemicals. Emerging research trends and future prospects of Raman spectroscopy in drug detection are also discussed in depth, including refining experimental protocols, enhancing analytical models, and improving the portability and accuracy of field detection equipment. Expanding the application scope of Raman spectroscopy for drug detection is also emphasized, particularly in supporting China’s public security forces in anti-narcotics efforts. This review aims to provide insights to guide the development of strategies and technologies in the fight against drug abuse and trafficking.
Abstract (
)
Download PDF
HTML
(
)
Knowledge map
Save
Forensic science serves as a crucial guarantee for safeguarding judicial justice and enhancing national crime governance capacity. The localization and quality development of its equipment directly affect the safety and efficiency of law enforcement operations in China. This paper analyzes practical equipment requirements within three key forensic domains: crime scene investigation and evidence collection, laboratory analysis and identification, and digital forensics and emerging crime governance. Currently, portable and intelligent equipment for crime scene investigation has advanced, supported by China's light industry foundation; laboratory analysis equipment is gradually closing the gap with international advanced levels; and digital forensics equipment for emerging crimes has rapidly developed driven by AI and internet technologies. However, challenges remain, including reliance on imported core technologies and an incomplete standardization system. Taking DNA equipment as an example, this paper proposes a high-quality development path: beginning with domestic reagents tailored to the genetic traits of the Chinese population, followed by progressive achievements in large-scale sequencer assembly, core component development, and independent analysis software creation. Simultaneously, a standardization system covering reference materials and methodologies should be established. Looking ahead, China's forensic equipment development should adhere to the principles of self-reliance, standardization, and practical applicability. Through deeper technology integration and AI-driven solutions, China can strengthen its advantages, enhance equipment localization, and address deficiencies. Collaborative efforts among police security institutions, enterprises, universities, and research institutions will promote the systematic adoption of domestic equipment. Furthermore, the Belt and Road Initiative can facilitate the export of equipment and spread of standards, ultimately enabling China to transition from following to leading in forensic science, thereby providing core technological support for judicial justice and modern crime governance.
Abstract (
)
Download PDF
HTML
(
)
Knowledge map
Save
The main characteristic of forensic science is that, with the goal of continuously improving the efficiency and accuracy of case handling, it is transformed and re-innovated through the introduction of theories, technologies and useful practices from various fields. Along with expanding, the scope and depth of forensic science research, more scientific and accurate methods are provided for the admissibility of evidence and the confirmation of facts in cases, as well as building a more systematic and complete system of knowledge and technology. In order to standardize the evaluation and management of technological innovation in this field, improve the efficiency of innovation and the stability and reliability of technology, and promote technological exchanges and cooperation as well as the optimal allocation of resources, it is attempted to introduce the evaluation of technological maturity in this field and to explore the applicable determination standards. The systematic advancement of related research is inferred by analyzing the three continents and their mixed communities, and the guiding role of technological maturity in the development of criminal technological innovation is analyzed by example.
Abstract (
)
Download PDF
HTML
(
)
Knowledge map
Save
In the investigation of telecom network fraud cases, law enforcement agencies frequently encounter a critical technical challenge. Crucial evidence such as fraudulent websites often becomes inaccessible due to the delay in reporting the case. Criminal organizations typically employ tactics including domain unbinding, IP address rotation, and server replacement to evade detection, thereby obstructing investigators from accessing webpage content and potential evidentiary clues, resulting in the loss of pivotal evidence. In response to this issue, this paper introduces the application methods of network asset mapping technology and related search engines based on two cases. First, conducting searches through keyword syntax based on shared characteristics and similar structures in webpage code. Second, resolve case-related domains to corresponding IP addresses and perform reverse lookups to identify newly bound domains under those IPs, and collect network intelligence through identifying homologous active webpages. This study aims to provide actionable references for investigating telecom network fraud cases.
Abstract (
)
Download PDF
HTML
(
)
Knowledge map
Save
In recent years, portable X-ray machine has gradually been applied in forensic autopsy to detect foreign objects, fractures, organ injuries, bleeding lesions, etc. It provides guidance for autopsy procedures, and offers technical support for diagnosing causes of death and analyzing injury mechanisms. This article introduces the application of using portable X-ray machine in virtual examination of corpse in a high fall death event. The advantages of this technology compared to traditional autopsy methods were discussed, and the key points, shortcomings, and risks were also analyzed. The article looks forward to the application prospects of this device in forensic virtual examination, aiming to provide case references and technical guidance for forensic virtual autopsy procedures.
Abstract (
)
Download PDF
HTML
(
)
Knowledge map
Save
Touch samples have become the most common type of biological evidence in theft cases, yet they remain challenging to analyze. Improving the effective detection and case matching rates of biological samples collected from theft scenes has long been a challenge in forensic DNA testing. In this study, 440 touch samples were randomly selected from theft cases handled by the forensic laboratory over the past year. These samples were analyzed using DNA quantification, followed by amplification optimization based on quantification results to investigate the impact of implementing DNA quantification protocols in such cases. Among the 440 samples, 171 yielded successful STR profiles (38.9% success rate) following re-optimized amplification based on DNA quantification data, representing a 23.7 percentage point improvement compared to suboptimal initial amplification results. Of these, 31 samples generated valid STR profiles, which matched 26 criminal cases. This study confirms that DNA quantification enhances both effective detection and case matching rates for touch samples in theft cases investigations. Furthermore, this approach shows potential for assisting in evidentiary attribution in forensic practice.
