Construction of feature analysis model for demeanor evidence investigation based on data mining algorithm

Demeanor evidence refers to the authoritative opinions of experts using analytical instruments to collect and interpret the formation of external physiological symptoms of the entire human body. With the development of modern technology, the observation of demeanor has entered a new stage of multimodal quantification. The quantifiable results of demeanor observation with the help of scientific and technological instruments play an important role in judicial work such as search, interrogation, and evidence review. The use of man-made reasoning in the field of legal settling has become increasingly broad. Specialists are giving increasingly more consideration to the securing of modular proof. This paper proposes the development of a component examination model for modular proof examination applications. The technique for this paper is to apply the guileless Bayes strategy, propose a superior information mining calculation, and lay out a model for proof observation and examination. The function of these methods is to systematically explore the basic theoretical issues of demeanor evidence based on the status quo of judicial application of demeanor evidence. Through the prediction of individual demeanor based on data mining algorithm, the evidence analysis model is designed, the neurobiological experiment is carried out, and the demeanor evidence animal stress model is constructed to verify the scientific basis of multidemeanor evidence observation. The experimental results showed that after repeated stimulation of SD rats, the maximum changes in the expression of HSP70 gene and SAA gene were 10.77 and 14.1 respectively, reflecting the high reliability of demeanor evidence biological experiments. The model can improve the accuracy of evidence use, and the correct use of demeanor evidence can realize the true litigation value of intelligent justice and the concept of human rights protection, and promote the construction of intelligent justice.


Introduction
At present, the understanding of demeanor evidence is limited to "assistance" in judging the authenticity of oral evidence at the trial stage.The discussion at the technical level should include both scientific connotation and technical extension.Demeanor evidence that can be scientifically observed and analyzed is the result of multimodal analysis, supported by current demeanor science and psychology.Demeanor evidence refers to a new, multimodal polygraph technology, which means that in order to prove the facts of the relevant case, professionals rely on analytical instruments to conduct the analysis of the overall external physiological expression of the human body that is not controlled by subjective will under the control of autonomic nerves.Concluding observations formed after analysis and interpretation.Demeanor evidence is a kind of scientific evidence.The judge of factual issues can judge the credibility of witnesses' testimony by their behavior when testifying in court.Multimodal interpretation of the subjects using eye movement detection technology, speech emotion analysis technology, cognitive EEG detection technology, physiological polygraph technology, facial microexpression recognition analysis technology and body behavior analysis technology [1].The multimodal demeanor analysis equipment can integrate all the behaviors of the test subjects, conduct an overall analysis on an analysis platform, eliminate the theoretical defects and operational errors that may exist in a single technology, and serve judicial practice applications with higher accuracy.Regarding the generation of demeanor evidence, people will have different understandings from different interests and perspectives.
Judging the authenticity of evidence by observing the state and changes of demeanor and obtaining evidence based on the observation of demeanor are widely used in the current investigation activities such as interrogation, questioning, search and identification in China.However, they are currently used only on the basis of work and life experience and intuition, and are not objectively evaluated and quantified on a large scale using technological means.The situation is similar at the prosecution and trial stages.In practice, judicial officers use demeanor observation to a certain extent to facilitate the evaluation of evidence, to corroborate the probative force of evidence, and to form judicial discretion.China already has increasingly specialized criminal evidence rules.With the continuous implementation of a series of judicial reform measures, the Supreme People's Court and other departments have successively promulgated some relatively important judicial interpretations.Some of these judicial interpretations involve the adjustment and development of criminal evidence rules.The motivation behind this paper is to examine the application worth and application rules of modular proof, in order to contribute to the development of China's criminal evidence system.The appearance of demeanor evidence can help judges accurately understand the evidence and assist relevant evidence to successfully complete the proving activities.
The correct application of demeanor evidence can realize the true value of litigation and the concept of human rights protection of intelligent justice, and 1 3 Construction of feature analysis model for demeanor evidence… promote the construction of intelligent justice.Otherwise, it will cause social ethics conflicts, rights violations and wrong conclusions.So we need to establish a more accurate model to improve the accuracy of modal evidence application.Therefore, he analysis model based on data mining constructed in this paper improves the original algorithm and puts forward new algorithm ideas.
Focusing on the issues existing in fluffy C-implies grouping calculation, this paper proposes to utilize information mining calculation to decide the underlying number of bunches, and loads the bunch place capacity and participation work.At the same time, this paper improves the data mining algorithm, reduces the influence of the selection of adjustment coefficient on clustering, and improves the practicability and feasibility of the algorithm.This paper finally proposes a sample-weighted fuzzy C-means clustering algorithm based on data mining and then conducts experimental verification and analysis.This article explores research issues that exist in the theory of demeanor evidence.Combined with the improvement of domestic and foreign researchers, this paper proposes an improvement for the calculation problem of focal element explosion in the calculation of combination rules, and compares and analyzes the combination results of numerical experiments and the evidence theory combination rule formula.Through the model test and the comparison with the experimental results of previous research methods, this paper provides a new idea for the data mining algorithm in the field of behavior evidence analysis and expands the application scope of the algorithm, which expands the application range of the algorithm, which enables demeanor evidence to carry out more accurate judicial work in a wider range.

Related work
Demeanor evidence, although never given formal evidentiary status, has always been considered an important basis for the court to judge the reliability of a witness.Systems and rules established on the basis of demeanor evidence, such as the jury system, in-court testimony, the right to confrontation, the hearsay rule, and crossexamination, play a crucial role in modern justice.The modern judicial system recognizes the existence and use of demeanor evidence in the laws or jurisprudence on plea bargaining, non-prosecution, sentencing recommendations, tainted witnesses, impeachment of discharged jurors, jury verdicts, witness competence and reliability, sentencing, evidentiary competence, and the first instance court in fact finding [2].Demeanor observation also plays a role in the fields of civil mediation, judicial correction, and risk assessment of persons serving sentences.

Literature affirming demeanor evidence from the perspective of scientific validity and rationality
Homegrown and unfamiliar establishments have completed a ton of examination and applications in the field of information mining characterization calculations.A detailed analysis of the existing treatises on the research on indicative evidence is necessary for the study of the rules for the use of criminal indicative evidence in China.For example, Prof. Zongzhi Long defines demeanor evidence as the nonverbal state of a witness when testifying, and points out that the court can analyze demeanor evidence to examine and judge the authenticity of the testimony, so demeanor evidence is an important "auxiliary evidence" in court [3] Prof. Qiwei Chen points out that demeanor evidence can help law executor to judge the authenticity and reliability of testimony in criminal prosecution practice, and can also be used as a clue for investigation and as evidence in special cases.However, it does not currently have a legal form of evidence.Through the analysis of demeanor information, it is found that it should belong to material evidence and has credibility, but its use should be restricted due to its immediacy, uncertainty, and multiple variability of interpretation [4] Prof. Yisheng Cai responded to the academic community's doubts about the legitimacy and rationality of demeanor evidence, arguing that although there are shortcomings compared to other scientific evidence, psychological and empirical studies in court can confirm that demeanor evidence has a certain degree of scientific validity and rationality, and has an irreplaceable role in judicial practice [5].Laurie L. Levenson has argued that the demeanor of various subjects both inside and outside the courtroom can affect the outcome of a trial, and there is no room for a trial to be purely free of demeanor [6].Numerous scientific experiments have shown that although the value of affective evidence should not be exaggerated, it is undesirable to simply deny its credibility [7].

Literature denying demeanor evidence from the perspective of traditional stereotype, traditional evidence rules and the free will of the observed person
After judicial practice developed from perceptual cognition to rational analysis, and from subjective judgment to objective quantitative analysis, demeanor evidence exists only in name in judicial practice, and its definition is controversial.At present, there are three main dilemmas in the application of demeanor evidence.First, due to the traditional stereotype of "psychological observation," judicial practitioners lack confidence in the effect of demeanor evidence.Secondly, due to the lack of understanding of the scientific basis of demeanor evidence and the limitations of traditional evidence rules, judicial practitioners affirmed the ability of demeanor evidence to prove the facts of the case, but denied the ability to produce evidence to prove the facts of the case.Third, it is controversial whether the acquisition of demeanor evidence violates the free will of the observed person.For example, "A Wipe of the Hands, A Lick of the Lips: The Validity of Demeanor Evidence in Assessing Witness Credibility" by Jeremy A. Blumenthal argues that assessing the credibility of witnesses at trial through demeanor observation has been deeply rooted in practice for almost three thousand years [8]."Demeanor Impeachment: Law and Tactics" by Prof. Imwinkelried Edward argues that only by clarifying the evidentiary status of demeanor observation can demeanor observation be rationally and effectively portrayed [9].Prof. Wellborn has argued in Cornell Law Review that demeanor evidence provides historical and modern justification for public trials,

3
Construction of feature analysis model for demeanor evidence… arguing that the jury system, in-court testimony, the right to confrontation, the hearsay rule, and cross-examination are all founded on demeanor evidence [10].Prof. Bruce Fraser, argues that for so-called machiavellian liars, because they do not have the moral constraints of ordinary people and do not feel remorse or guilt; therefore, they often do not have the demeanor expression of a lie so that they are lack the possibility of being identified [11].

Deficiencies in existing studies
Most of the existing studies focus on the theoretical aspects such as generative and cognitive principles, historical investigation, legitimacy, rationality, and comparative research at home and abroad, and certain results have been achieved.However, there are still problems in existing study literature such as lacks the exploration of the specific application rules, legislative construction, judicial application, and risk prevention of demeanor evidence.There are few literatures about the risk and prevention of the application of demeanor evidence from a technical perspective.In addition, the above research results mentioned the advantages and disadvantages of demeanor evidence, but they were not studied in actual cases, which lacked certain reliability.Therefore, with the objective to make a breakthrough from this point, the author carried out analysis sorting existing literature and judicial documents, and attempted to apply the guileless Bayes strategy, propose a superior information mining calculation, and lay out a model for proof observation and examination, so as to improve the effectiveness of the application of demeanor evidence.This study is expected to improve the accuracy of the application of demeanor evidence and maintain social safety and order.
3 Reconnaissance methods of data mining algorithms

Naive Bayesian methods
The more important technique in data mining is the classification technique.In the entire course of data mining, it is vital to characterize the information sensibly.Designing an improved algorithm of data mining can further improve the efficiency of existing data mining algorithms and improve the results of data mining.The legitimacy of characterization results straightforwardly influences the exactness of data mining results [12].The development strategies for classifiers mostly incorporate the accompanying: choice tree grouping strategy, numerical measurements technique, man-made consciousness technique, neural organization strategy, and so on as per different examination headings of order calculation, it tends to be partitioned into the accompanying classifications: choice tree arrangement calculation, Bayesian grouping calculation, affiliation rule, neural organization, k-closest neighbor calculation, hereditary calculation, unpleasant set, and so on [13].This part first introduces the concept and theoretical techniques of Bayesian method, and then proposes a new incremental decision tree algorithm combining Bayesian method and decision tree algorithm.It solves the problem of incremental learning of data in the process of data mining classification.This article presents the basic theory and working principle of Bayesian processing, and gives an example of Bayesian processing.Allow x to indicate an information test whose class name is obscure.In the event that the information test x has a place with a specific class D, let H be the theory.This paper desires to track down P(H|X), which is communicated as the probability that speculation H holds for the example information x.The computation strategy for P(H|X) is as per the following: where P(H) is known as the prior probability, or the prior probability of H. P(x|H) addresses the probability that x can be seen when the H condition is valid.P(H|x) addresses the posterior probability accepting that the condition h holds under condition x.
The essential strides of the Bayesian arrangement process are as per the following: (1) The n-layered element vector X = x 1 , x 2 … x n is utilized to address all informa- tion tests, and the n measurements of the examples with n ascribes ple information without a class mark, and the classifier predicts that X has a place with the class with the most elevated posterior probability (under the condition X).That is, the Bayesian separator will characterize the obscure example information into the class Ci (1 < i < m) if and provided that The most extreme deduced is thought to be the class Ci with the biggest P C i |X , and P C i |X not entirely settled by the accompanying Bayes hypothesis: (3) Since P(X) is a consistent, the bigger the worth of P(X|C i )P C j , the bigger the worth of P C i |X .The prior probability of class C is obscure, then, at that point, it very well might be expected that these class probabilities are equivalent, to be specific P C 1 = P C 2 = … = P C m , so the issue turns into the expansion of DD.P(X|C i ) is regularly alluded to as the probability of information X given Ci, while the theory Ci that expands P P(X|C i ) is known as the high probability speculation.In any case it needs to amplify P(X|C i )P C j .Note that the supposi- tions are not equi-plausible, so the prior probability of a class can be determined with P C i = s i ∕s , where s i is the quantity of preparing tests in class Ci and s is the all out number of preparing tests.(4) Due to the huge number of qualities in the informational index, the computation will create a great deal of additional upward.To diminish the additional expense (1) P(H|X) = P(X|H)P(H) P(X) . (2) Construction of feature analysis model for demeanor evidence… of figuring P(X|C i ) , class condition autonomy can be accepted.Since the class mark of the example is known, and as per the presumption that the traits are restrictively autonomous of one another, it tends to be reasoned that the properties do not rely upon one another, so: The worth of P X 1 |C i , P X 2 |C i … P(X k |C i ) can be created by learning the preparation tests.In the event that A k is a discrete trait, P X k |C i = s ik ∕s i , where s ik is the quantity of preparing tests of class Ci with esteem x k on property A k , and s i addresses the size of the preparation tests in Ci.
Assuming that A k is a consistent esteemed property, it is by and large accepted that the characteristic complies with a Gaussian appropriation, for example where g X k , u C i , Q C i is the Gaussian dissemination capacity, and u C i and Q C i are the mean and standard deviation, separately.A Bayesian way to deal with assessing probabilities can be utilized, characterized as follows: In this formula, nc is the same as the definition, P is a certain a priori estimated probability, the value of m is equal to the size of the current sample, this value is mainly used to use the observed data to measure P [14].This formula can be understood as follows: the size of the original n actual samples is enlarged, and m virtual samples are added, and these samples are distributed according to P. In the absence of other information, assuming that the priors are uniform is an efficient way to choose P. That is, if an attribute has k possible values, then set P = 1/k.In order to estimate P(age < = 30lBuys_computer = "Yes"), suppose the attribute age has three possible values, so the uniform prior probability can be set to P = 0.33.If m = 0, the m-estimation is equivalent to the simple scale nc/n.If both n and m are nonzero, the prior probability P and the observed ratio nc/n can be combined according to the weight ratio m.

Improved data mining algorithm
Enormous information objects have the intricacy of the condition of information spatial dispersion, for example, the circulation examples of information objects of various sizes, shapes and densities.To successfully find the impartially existing dispersion examples of mind boggling morphological information objects in information cross-examination, it is important to utilize productive quality weighting and thickness grouping calculations.It computes the circulation thickness of information objects in the information space, decides the thickness of the thickness fascination (3) focuses (outrageous focuses) and the thickness of the information objects to the thickness fascination focuses.Along these lines, the powerful grouping of bunches of various sizes, shapes and densities can be understood, and afterward the viable mining and investigation of a lot of information can be acknowledged [15].When the adjustment coefficient is too large, the data clustering effect of the original data mining algorithm will be relatively unsatisfactory, and will directly affect the clustering effect.In practical applications, the selection of the adjustment coefficient is too large, which will often directly affect the final data mining results.It is necessary to improve the algorithm to reduce the effect of the adjustment coefficient selection on the clustering, and to improve the practicability and feasibility of the algorithm.
In the process of data processing, the selection of the density parameter value usually represents whether the location of the density attraction point is the most dense location of the data objects [16].As shown in Fig. 1, it is the clustering result of using the global density calculation method when the value of Q is different.
As can be seen from the figure, the selection of different density parameter values is different, and the location of the density attraction point will also be different.The better CADD calculation proposed in this subject is not reasonable for worldwide thickness estimation.In the bunching system of the superior calculation proposed in this paper, it is important to leave the worldwide neighborhood range utilized by the first calculation and utilize the unique neighborhood sweep.In particular, its numerical depiction recipe.
The unique neighborhood range versatile thickness reachable distance is characterized as: In the formula, R is the reachable distance of the underlying thickness, and Ai and Ai + 1 separately address the thickness upsides of the two bunches thickness fascination focuses decided progressively.
Information direct thickness alludes toward the general thickness of the information space, and furthermore by and large alludes to the amount of the impact Fig. 1 Clustering results of the improved algorithm when the density parameter Q is 5 and 2

3
Construction of feature analysis model for demeanor evidence… elements of all elements in the wake of displaying.The formula can be communicated as: In the formula, the Gaussian capacity of the right formula addresses the impact of every important informative element on the Xi point, and Q is the thickness boundary.The thickness reachable distance alludes to any information object X in the information bunch space, and the distance R between the information.The round region with an information as the middle and the information distance as the sweep compares to the reachable thickness distance field of the information object.
In the formula, coefR alludes to the change coefficient of the distance, and the coefficient esteem is more prominent than 0 and under 1. mean(D) alludes to the normal distance of all information objects.D is an article assortment of information.
As indicated by the meaning of dynamic neighborhood span, as per the versatile thickness reachable distance of every fascination point, the nearby re-parcel of the dataset is done [17].On the off chance that during the fractional division of the informational collection, nearby additions ΔD1, ΔD2, and ΔD3 show up after the informational index is separated, it is important to re-search the information focuses that can arrive at each roundabout thickness in the gradual information.Thusly, dynamic gradual grouping of the sectioned dataset is understood, as displayed in Fig. 2. The following figure shows that after segmenting the data set, it does not undergo fission, that is, the number is the same, C1 is 13, C2 is 25 and C3 is 12.
Clustering feature (CF) is used to describe the information summary of subclustering of objects, and includes the cluster information projected onto other attribute sets, and its formula is: In the formula, N represents the number of all tuples in the subcluster.Assuming C X = t 1 , t 2 … t n , the associated cluster feature (ACF) formula is as follows: Then we perform cluster density evaluation.Let S[X] be the set of N datasets t 1 , t 2 … t n projected onto attribute set X. Then the distance metric formula of S[X] is: In the formula, represents the distance measure between tuples.When the distance of S[x] is larger, the deviation of its dataset projected to attribute set X is larger.
In the plan of this better calculation, it can evaluate the scope of qualities, or at least, track down the most extreme and least upsides of characteristics, and the computation strategy is straightforward.For instance, on the off chance that the worth is planned to [0,1], the computation formula is as per the following:

Evidence detection and analysis model
The proof surveillance model proposed in this paper puts the legal cycle in three-layered space, and forms the relationship between forensic strategies, forensic requirements, and time constraints.Forensics strategy and forensics monitoring are divided into data layers to ensure the availability of evidence and supervise the entire evidence collection process [18,19].This process is shown in Fig. 3.
The multidimensional computer forensics model includes steps such as digital forensics, forensic preparation, monitoring of the entire forensic process, physical forensics, evidence presentation and summarization [20].Evidence analysis is a core and important step in collecting criminal evidence.Since the raw data is huge, noisy, incomplete, ambiguous and random, this paper extracts hidden, beforehand obscure and possibly helpful data and information.Further forensic analysis should be accessible and should provide readable, practical and legally binding evidence.The combination of optimization and data mining algorithms can improve the performance of evidence analysis.
Principal component analysis (PCA) is a statistical technique that replaces the original measure with some representative synthetic measure.Direct blends of the first P measurements are frequently utilized as new composite measurements, ( 9) Construction of feature analysis model for demeanor evidence… normally addressed by the change of the primary straight mix F1 picked [21].In the event that the second straight blend F2 is chosen, the data currently in F1 would not be shown in F2.In other words, Cov(F1, F2) = 0 , F2 is known as the second head part, and furthermore comprises the third to p head parts.The principle steps of head part investigation are.Normalization of test pointer information.Let n tests, p pointers, get the first network: In order to standardize the sample data, the mean and variance of the sample data should be obtained.Normalized data matrix Y.That is, it performs a normalized transformation on each index component.The conversion formula is in the form: The sample mean is ȳj = 1 n ∑ n k=1 y ki and the sample standard deviation is Get the normalized data matrix: be the ment pace of the Ith head part.The bigger the worth, the more grounded the capacity of the Ith head part to incorporate data, that is: Practically speaking, a rate is generally determined to choose the number of head parts to keep.Along these lines, high-layered information is changed over into lowlayered information, which can be effortlessly characterized by bunch examination [22].
Set the example space X = x 1 , x 2 … x n , partition X into K classes, and utilize fluffy lattice U = u ij to address the enrollment level of the Ith test to the jth class.It is characterized as follows: ∑ m j=1 i Construction of feature analysis model for demeanor evidence… where c j (j = 1, 2 … K) is the cluster center, d ij = x i − c j is the gap between x i and c j , and the Euclidean distance in the Euclidean space.b is the fuzzy index, that regulates how soft or fuzzy the clustering process is.In contrast to hard clustering, which assigns each data point to a single cluster, fuzzy clustering assigns each cluster center a degree of membership.The clustering algorithm (data mining algorithm) level of fuzziness is determined by "b" and its The procedure becomes the hard clustering algorithm when "b" is set to 1, and each data point is assigned to a single cluster.The clustering process becomes more fuzzy as "b" rises, allowing for partial membership of data points in several groups as well as overlapping clusters.
In general, there are a number of ways to figure out what "b" should be in the fuzzy clustering process.The clustering algorithm's performance is assessed using a set of performance measures, such as the silhouette score or the F-measure, in a method known as trial and error, where alternative values of "b" are tested on the dataset.Evidence analysis is the core process in the process of extracting, analyzing and submitting computer forensic evidence.How to extract suspicious intrusion data from massive information is a hot issue in forensic analysis and intrusion detection research.Through the long-term study of computer crime forensics technology and the in-depth study of data mining technology, this paper proposes a coordinated PC legal sciences framework model in light of host and network and the use of information mining calculation in wrongdoing criminology [23].

Design of evidence analysis model
In this experiment, neurobiological experiments were carried out, a demeanor evidence-bearing animal stress model was constructed, and the molecular biological function of biological demeanor evidence was analyzed to verify the scientific basis of multimodal evidence observation.Secondly, conduct demeanor psychology experiments to form a comprehensive analysis of speech, physiology, eye movements, facial microexpressions, body movements, etc., to verify the entire process of behavior and its relationship with external stimuli.Finally, a model is established based on the experimental data, and the accuracy of the multimodal evidence is calculated by the Bayesian algorithm, which proves the authenticity, relevance and accuracy of the demeanor evidence in the field of evidence law [24,25].
From biology, demeanor is a general term for the expression and activation of neurotransmitters under the action of stressors.It maintains internal stability by inducing the immune system and the endocrine system through the nervous system.Investigate signaling cascades and molecular mechanisms in the human body when demeanor evidence is generated.In the experiments, demeanor generation was induced by simulating exposure to a combination of acousto-electric stimulation.For the acousto-electric stimulation, 25 healthy rats were randomly divided into five groups, with each group containing 5 individuals [26,27].After stimulation, hemocytes were collected from 5 individuals of each group at 0 h, 1 h, 3 h, 12 h and 24 h, respectively.The hemocytes collected with 0.109 mol/L sodium citrate from each group were centrifuged at 1000 × g for 10 min at 4 °C, then archived in TRIzol and stored in − 80 °C refrigerator until RNA extraction.
Total RNA was extracted from the collected samples using TRIzol Reagent according to the manufacturer's protocol.RNA was dissolved in DEPC-treated water, and the integrity of the RNA was verified by electrophoresis with a 1.0% agarose gel, and then the concentration and purity were examined at 260/230 and 260/280 absorbance ratios using a Nanodrop 2000c spectrophotometer.Then, reverse transcription (RT) was carried out to prepare the cDNA template using a PrimeScript™ RT reagent Kit with gDNA Eraser (TaKaRa, Japan) according to the manufacturer's instructions.
In order to test the correlation of demeanor evidence from a scientific point of view, this experiment conducted biological experiments in SD rats targeting the HSP70(Genbank: XM_039085314) and SAA (Genbank: NM_001009478) mRNA of the endocrine and immune systems.Gene-specific primers of RatSAA F: 5′-AGA AGG TTC CAG AAG CCC AAAT-3′; RatSAA R: 5′-CAG AGG ATA ATG ACG GTG ACAAA-3′ for the SAA and RatHSP70 F:-5′ AGC AAC AGC CAC AGA CAC CAG-3′, RatHSP70 R: 5′-ATT CCG TTC CTT CTC CAG TTT ATC -3′ for HSP70, for qRT-PCR were designed to amplify an approximately 200 bp fragment.Each assay was performed in triplicate with Actin (Genbank: EF156276) mRNA with RatAct F: 5′-GTT GTC CCT GTA TGC CTC TGGTC-3′, RatAct R: 5′-ATG TCA CGC ACG ATT TCC CTC-3′ primers and as an internal control.Then, qRT-PCR was conducted using a LightCycler 480 (Roche, Switzerland) and 2 × SYBR green mix (DSBIO, China) according to the manufacturer's protocol.The analysis of the melting curve was performed to confirm the specificity at the end of q-PCR.The relative expression of SAA was calculated using 2 −ΔΔCT method.The recorded experimental data are shown in Figs. 4 and 5.
The expression of HSP70 and SAA mRNA were up-regulated 10.77 and 14.1, fold compared with the control, respectively.The demeanor evidence of rats in this experiment is more obvious, indicating that the model prepared in this experiment is successful; meanwhile, the generation of demeanor evidence was based on the biology molecules.The analysis shows that the demeanor evidence based on this experiment has a fairly stable correlation that can be measured by reliability criteria.Molecular changes in biological experiments are beneficial to help build demeanor evidence investigation feature analysis models and expand ideas for modal evidence collection methods.In this experiment, SD rats finally showed a stable physiological response under repeated stimulation with the same content and intensity, reflecting the highly reliable demeanor evidence of biological experiments [28].The stable results of biological experiments provide strong experimental support for the correlation assessment of demeanor evidence.
Furthermore, in demeanor facts, the verbal content not only describes the data of the physiological expression, but only describes some of the information conveyed by the declarant.Therefore, demeanor analysis should be an analysis of the overall response of the observed subject's adaptation to a specific stimulus [29].This paper observes and analyzes all demeanor information by integrating the subject's eye movements, voice, subtle facial expressions, physiology, and other information, resulting in a more realistic and objective performance in terms of accuracy 1 3 Construction of feature analysis model for demeanor evidence… Fig. 4 Expression of HSP70 mRNA during demeanor generation in SD rat model Fig. 5 of SAA mRNA during demeanor generation in SD rat model and reliability.Facial expression recognition and analysis experts say that the facial activity coding system (FACS) based on facial expression recognition technology can achieve a success rate of more than 70 percent in polygraph detection.The current demeanor analysis tools, supported by precision cameras and analytical instruments, can capture the microscopic features of the investigated object at high speed, and complete microscopic feature capture and feature calculation every 24 ms [30,31].The latest demeanor analysis systems developed based on these theories and methods have greatly improved the accuracy of the analysis.For example, the intelligent judgment system jointly developed by the Dalian People's Procuratorate and related technology companies has passed the actual verification, and the system accuracy rate is stable at 9 points (out of 10 points).

Demeanor evidence and polygraph evidence
Polygraph evidence is one that is often used in criminal investigation procedures.It needs to conduct some professional tests on the relevant personnel of the case with the help of test media such as a polygraph, and then integrate and analyze the test data to draw a conclusion whether the relevant personnel of the case are lying.Polygraph tools such as polygraphs are an indispensable medium for drawing conclusions of polygraphs, and what they test is not the lies told by the test subjects.And they are based on the physiological changes produced by psychological fluctuations when they lie [32].The main function of the polygraph is to record the parameters generated by these changes, and the investigators can obtain the final conclusion on whether the tested object is lying or not through the analysis of the parameters.
China's Criminal Procedure Law stipulates that evidence is important and can be used to prove the facts of a case.However, there is no corresponding category of demeanor evidence in Chinese laws and regulations.Then the experiment proves that the model can describe user demeanor intuitively and accurately by analyzing actual network data and simulation experiments, and accurately distinguish malicious users and risk users [33].
Demeanor evidence, which has not received enough attention in academia, "exists in judicial practice in thousands of ways."In order to fully grasp the judicial workers' cognition of the extension scope of demeanor evidence in practice, the author conducted questionnaire surveys and interviews among criminal judicial practitioners such as public security organs, procuratorates, courts and criminal courts, and criminal defense lawyers.On the basis of clarifying and explaining the basis of physiological science, please enumerate the range of demeanor evidence that they have felt or used in business practice as much as possible.After summarizing, classifying and analyzing 853 responses, it is concluded that judicial practitioners enumerate the main extension of demeanor evidence as follows, ranked according to the number of occurrences, that is, the degree of recognition: First, body movements.Respondents in a total of 671 questionnaires believed that demeanor evidence included body movements, accounting for 78.6% of the total number of questionnaires.Second, facial expressions.Respondents to a total of 631 questionnaires believed that demeanor evidence included facial expressions, 1 3 Construction of feature analysis model for demeanor evidence… accounting for 73.9% of the total number of questionnaires.Third, the physiological response category.Respondents to a total of 625 questionnaires believed that demeanor evidence contained physiological responses, accounting for 73.2% of the total number of questionnaires.Fourth, voice and intonation.Respondents in a total of 335 questionnaires believed that demeanor evidence included voice intonation, accounting for 39.2% of the total number of questionnaires.Fifth, eye movements.In addition, there were 14 respondents who expressed disapproval of demeanor evidence, accounting for 1.6% of the total number of questionnaires.The specific statistics are shown in Fig. 6.
By analyzing the questionnaire survey data, two aspects of the practical application of the extension scope of demeanor evidence can be summarized.First of all, the vast majority of criminal justice workers have perception and experience of the application of demeanor evidence in practice, and only a very small number (1.6%) of the respondents said that they have never experienced the practical application of demeanor evidence.Embodies the "mass base" that is widely used in demeanor evidence.Second, the current criminal justice workers have a certain degree of overlap in the enumeration of the extension of demeanor evidence.In legal education and academic research, where there is a relative lack of research on the concept of demeanor evidence, this enumeration overlap is clearly derived from judicial practice work and daily life experience.This conclusion reflects the "common understanding" of judicial practitioners on the extension of demeanor evidence.

Establishment of multimodal calculation model
In the multimodal modal analysis, the existing data and core indicators are used to establish a multidimensional statistical classifier, and high-definition cameras and sound acquisition equipment are used to collect high-definition images of the Fig. 6 Extension enumeration of demeanor evidence in judicial practice observed whole body action pictures and facial expressions.These are loaded with audio and video data.In the behavior analysis system of another control room workstation, the body movements of the observed person are encoded, marked and analyzed, and then transmitted to the control room workstation through the local area network, and synchronized with the audio and video data into the behavior analysis system.The behavior analysis platform performs frame-by-frame analysis of eye movement behavior, facial expressions, voices, emotions, and collected physiological data.This behavioral science and technology approach can understand the overall occurrence of observed behavior and its behavior from several different perspectives.It is related to external stimuli, grasps the behavior and psychological state of the observed as a whole, analyzes the influencing factors in detail, excludes invalid data, so as to conduct more accurate observation and analysis, provide behavioral analysis conclusions and behavioral scientific suggestions for interrogation, and find breakthroughs in interrogation and screening.Help with the authenticity of the confession.
In order to verify the accuracy of the modal evidence under multimodality, the authors performed calculations with experimental data.Among the 196 test samples, 98 were real criminal suspects and 98 were in the control group.The five tests of eye movement, speech, physiology, EEG, and microexpressions were carried out in the order of the experiments, and the accuracy rates were 90% [34], 70% [35], 95% [36], 75.4%, and 90%.For the control group, the accuracies were 95%, 76%, 91%, 98%, and 80%.
Let A i be the event of "the ith detection result is true," and i ≤ 5 .B be the event of "the detected person is a criminal suspect."On this basis, when the second detection is performed, the probability of the event "the second detection result is true under the condition that the detected person is a criminal suspect" is shown in formula (22): On this basis, when the third test is performed, the probability of the event "the third test result is true under the condition that the detected person is a criminal suspect" is shown in formula ( 23): On this basis, when the fourth test is performed, the probability of the event "the result of the fourth test is true under the condition that the detected person is a criminal suspect" is shown in formula (24): It should be noted that, in the application of specific cases, not all six detection indicators can be collected from the modal observation of each case.Eye movement metrics cannot be collected when refusing to look at the screen.Of course, when the observed person has severe strabismus, severe stuttering, facial twitching, and even language barriers and visual impairments, there are even more conditions for observation.Therefore, specific to an observed person, there may be only one or several indicators available for observation, and the collection of all six indicators is only the "optimal" state, rather than the abnormal state.According to the above calculations, it can be seen that the more indicators collected, the higher the accuracy of multimodal modal observation.When the third indicator is integrated, the accuracy of the detection results can reach 99%.When the fourth index is integrated, the accuracy of the detection results can reach 99.9%.

Statistical analysis
analysis was performed with Statistical package for the social sciences 17.0(SPSS17).All data were presented as the mean standard deviation (M ± SD) throughout the text.Comparisons between two groups of samples were performed with Student's t-tests, and comparisons among more than two groups were performed with one-way ANOVA followed by Turkey's or Dunn's post hoc test.Values *P < 0.05 was considered significant, and **P < 0.01 was considered highly significant.

Discussion
This experiment attempts to infer the behavior patterns of criminals by analyzing the demeanor evidence left at the crime scene and combining traces of other relevant physical evidence.Due to the limitation of time and its own level, the research of the current paper is not deep enough.The development of modern technology has created the possibility of observing demeanor and obtaining demeanor evidence through scientific methods.The current criminal methods and anti-investigation skills are constantly being upgraded, which makes it more difficult for the court to trace the facts of the case, while the demeanor evidence that can be "understood" by modern technology has enriched the current methods for judging the authenticity of testimonies and confessions, and supplemented the way of the facts of the case has supplemented the means of confirming the crime or not, helping the investigators to find more clues for the investigation, assisting the court to "understand the crime" more comprehensively and objectively, convicting and sentencing.

Conclusion
This paper accepts characteristic proof as the examination item, and plans to talk about the current circumstance and application rules of criminal demonstrative proof with regards to the meaningful preliminary and the time of shrewd courts.It provides a Chinese version of the evidence rules for showing evidence in the court of criminal proceedings in an all-round way and avoiding the falsification of court trials.After far reaching investigation, the accompanying resolutions can be drawn: Under the background of current demeanor collection technology, demeanor evidence can be carried out simultaneously in the investigation process such as interrogation, on-site investigation, search, and identification, and in the trial.While obtaining more information related to the statement and case facts, it does not increase the cost of investigation time, and the non-contact demeanor observation and collection instrument does not infringe on personal and free will under the condition of informed consent of the observed.During the trial, witnesses testify in court can be regarded as the knowledge and approval of the court's observation of their demeanor, and there will be no additional cost and risk.However, the court can obtain richer information through demeanor evidence to review and judge the authenticity of the confession and the facts of the case.Under the constraints of relevant rules, the rational use of demeanor evidence can improve litigation efficiency, save litigation costs, reduce litigation waste, and realize the true efficiency value of litigation.Subjective observation or over-reliance on demeanor observation techniques and erroneous reasoning can create risks such as misjudged cases and imbalance between prosecution and defense, which will undermine substantive and procedural justice.It is necessary to regulate the procedural conditions for the application of demeanor observation in the legislation and formulate rules to prohibit the reverse use of demeanor observation in order to avoid the risks.Due to the time problem, the research on the extraction of specific information and the occurrence time determination of behavior time is not deep enough.At the same time, it is necessary to infiltrate the improved data mining technology into other fields.

Fig. 3
Fig. 3 Multidimensional computer forensics model diagram According to the experimental results, st.P(B) = 0.5 ,P(B) = 0.5, P(A 1 |B ) = 0.9 , P(A 2 |B ) = 0.7 , P(A 3 |B ) = 0.95 , P(A 4 |B ) = 0.754 , P(A 5 |B ) = 0.9 .P(A 1 According to the Bayesian principle, the probability of the event "the first detection result under the condition that the detected person is a criminal suspect" is true as shown in formula (21):

3
Construction of feature analysis model for demeanor evidence…