Evolutionary Decision-Making and Planning for Autonomous Driving Based on Safe and Rational Exploration and Exploitation

Decision-making and motion planning are extremely important in autonomous driving to ensure safe driving in a real-world environment.This study proposes an online evolutionary decision-making and motion planning framework for autonomous driving based on a hybrid data-and model-driven method.First,a data-driven decision-making module based on deep reinforcement learning(DRL)is developed to pursue a rational driving performance as much as possible.Then,model predictive control(MPC)is employed to execute both longitudinal and lateral motion planning tasks.Multiple constraints are defined according to the vehicle’s physical limit to meet the driving task requirements.Finally,two principles of safety and rationality for the self-evolution of autonomous driving are proposed.A motion envelope is established and embedded into a rational exploration and exploitation scheme,which filters out unreasonable experiences by masking unsafe actions so as to collect high-quality training data for the DRL agent.Experiments with a high-fidelity vehicle model and MATLAB/Simulink co-simulation environment are conducted,and the results show that the proposed online-evolution framework is able to generate safer,more rational,and more efficient driving action in a real-world environment.

Toward Trustworthy Decision-Making for Autonomous Vehicles:A Robust Reinforcement Learning Approach with Safety Guarantees

While autonomous vehicles are vital components of intelligent transportation systems,ensuring the trustworthiness of decision-making remains a substantial challenge in realizing autonomous driving.Therefore,we present a novel robust reinforcement learning approach with safety guarantees to attain trustworthy decision-making for autonomous vehicles.The proposed technique ensures decision trustworthiness in terms of policy robustness and collision safety.Specifically,an adversary model is learned online to simulate the worst-case uncertainty by approximating the optimal adversarial perturbations on the observed states and environmental dynamics.In addition,an adversarial robust actor-critic algorithm is developed to enable the agent to learn robust policies against perturbations in observations and dynamics.Moreover,we devise a safety mask to guarantee the collision safety of the autonomous driving agent during both the training and testing processes using an interpretable knowledge model known as the Responsibility-Sensitive Safety Model.Finally,the proposed approach is evaluated through both simulations and experiments.These results indicate that the autonomous driving agent can make trustworthy decisions and drastically reduce the number of collisions through robust safety policies.

Orientation and Decision-Making for Soccer Based on Sports Analytics and AI:A Systematic Review

Due to ever-growing soccer data collection approaches and progressing artificial intelligence(AI) methods, soccer analysis, evaluation, and decision-making have received increasing interest from not only the professional sports analytics realm but also the academic AI research community. AI brings gamechanging approaches for soccer analytics where soccer has been a typical benchmark for AI research. The combination has been an emerging topic. In this paper, soccer match analytics are taken as a complete observation-orientation-decision-action(OODA) loop.In addition, as in AI frameworks such as that for reinforcement learning, interacting with a virtual environment enables an evolving model. Therefore, both soccer analytics in the real world and virtual domains are discussed. With the intersection of the OODA loop and the real-virtual domains, available soccer data, including event and tracking data, and diverse orientation and decisionmaking models for both real-world and virtual soccer matches are comprehensively reviewed. Finally, some promising directions in this interdisciplinary area are pointed out. It is claimed that paradigms for both professional sports analytics and AI research could be combined. Moreover, it is quite promising to bridge the gap between the real and virtual domains for soccer match analysis and decision-making.

Ethical Decision-Making Framework Based on Incremental ILP Considering Conflicts

Humans are experiencing the inclusion of artificial agents in their lives,such as unmanned vehicles,service robots,voice assistants,and intelligent medical care.If the artificial agents cannot align with social values or make ethical decisions,they may not meet the expectations of humans.Traditionally,an ethical decision-making framework is constructed by rule-based or statistical approaches.In this paper,we propose an ethical decision-making framework based on incremental ILP(Inductive Logic Programming),which can overcome the brittleness of rule-based approaches and little interpretability of statistical approaches.As the current incremental ILP makes it difficult to solve conflicts,we propose a novel ethical decision-making framework considering conflicts in this paper,which adopts our proposed incremental ILP system.The framework consists of two processes:the learning process and the deduction process.The first process records bottom clauses with their score functions and learns rules guided by the entailment and the score function.The second process obtains an ethical decision based on the rules.In an ethical scenario about chatbots for teenagers’mental health,we verify that our framework can learn ethical rules and make ethical decisions.Besides,we extract incremental ILP from the framework and compare it with the state-of-the-art ILP systems based on ASP(Answer Set Programming)focusing on conflict resolution.The results of comparisons show that our proposed system can generate better-quality rules than most other systems.

Stroke Risk Assessment Decision-Making Using a Machine Learning Model:Logistic-AdaBoost

Stroke is a chronic cerebrovascular disease that carries a high risk.Stroke risk assessment is of great significance in preventing,reversing and reducing the spread and the health hazards caused by stroke.Aiming to objectively predict and identify strokes,this paper proposes a new stroke risk assessment decision-making model named Logistic-AdaBoost(Logistic-AB)based on machine learning.First,the categorical boosting(CatBoost)method is used to perform feature selection for all features of stroke,and 8 main features are selected to form a new index evaluation system to predict the risk of stroke.Second,the borderline synthetic minority oversampling technique(SMOTE)algorithm is applied to transform the unbalanced stroke dataset into a balanced dataset.Finally,the stroke risk assessment decision-makingmodel Logistic-AB is constructed,and the overall prediction performance of this new model is evaluated by comparing it with ten other similar models.The comparison results show that the new model proposed in this paper performs better than the two single algorithms(logistic regression and AdaBoost)on the four indicators of recall,precision,F1 score,and accuracy,and the overall performance of the proposed model is better than that of common machine learning algorithms.The Logistic-AB model presented in this paper can more accurately predict patients’stroke risk.

The Spherical q-Linear Diophantine Fuzzy Multiple-Criteria Group Decision-Making Based on Differential Measure

Spherical q-linearDiophantine fuzzy sets(Sq-LDFSs)provedmore effective for handling uncertainty and vagueness in multi-criteria decision-making(MADM).It does not only cover the data in two variable parameters but is also beneficial for three parametric data.By Pythagorean fuzzy sets,the difference is calculated only between two parameters(membership and non-membership).According to human thoughts,fuzzy data can be found in three parameters(membership uncertainty,and non-membership).So,to make a compromise decision,comparing Sq-LDFSs is essential.Existing measures of different fuzzy sets do,however,can have several flaws that can lead to counterintuitive results.For instance,they treat any increase or decrease in the membership degree as the same as the non-membership degree because the uncertainty does not change,even though each parameter has a different implication.In the Sq-LDFSs comparison,this research develops the differentialmeasure(DFM).Themain goal of the DFM is to cover the unfair arguments that come from treating different types of FSs opposing criteria equally.Due to their relative positions in the attribute space and the similarity of their membership and non-membership degrees,two Sq-LDFSs formthis preference connectionwhen the uncertainty remains same in both sets.According to the degree of superiority or inferiority,two Sq-LDFSs are shown as identical,equivalent,superior,or inferior over one another.The suggested DFM’s fundamental characteristics are provided.Based on the newly developed DFM,a unique approach tomultiple criterion group decision-making is offered.Our suggestedmethod verifies the novel way of calculating the expert weights for Sq-LDFSS as in PFSs.Our proposed technique in three parameters is applied to evaluate solid-state drives and choose the optimum photovoltaic cell in two applications by taking uncertainty parameter zero.The method’s applicability and validity shown by the findings are contrasted with those obtained using various other existing approaches.To assess its stability and usefulness,a sensitivity analysis is done.

A NovelMethod for Determining Tourism Carrying Capacity in a Decision-Making Context Using q−Rung Orthopair Fuzzy Hypersoft Environment

Tourism is a popular activity that allows individuals to escape their daily routines and explore new destinations for various reasons,including leisure,pleasure,or business.A recent study has proposed a unique mathematical concept called a q−Rung orthopair fuzzy hypersoft set(q−ROFHS)to enhance the formal representation of human thought processes and evaluate tourism carrying capacity.This approach can capture the imprecision and ambiguity often present in human perception.With the advanced mathematical tools in this field,the study has also incorporated the Einstein aggregation operator and score function into the q−ROFHS values to supportmultiattribute decision-making algorithms.By implementing this technique,effective plans can be developed for social and economic development while avoiding detrimental effects such as overcrowding or environmental damage caused by tourism.A case study of selected tourism carrying capacity will demonstrate the proposed methodology.

UAV maneuvering decision-making algorithm based on deep reinforcement learning under the guidance of expert experience

Autonomous umanned aerial vehicle(UAV) manipulation is necessary for the defense department to execute tactical missions given by commanders in the future unmanned battlefield. A large amount of research has been devoted to improving the autonomous decision-making ability of UAV in an interactive environment, where finding the optimal maneuvering decisionmaking policy became one of the key issues for enabling the intelligence of UAV. In this paper, we propose a maneuvering decision-making algorithm for autonomous air-delivery based on deep reinforcement learning under the guidance of expert experience. Specifically, we refine the guidance towards area and guidance towards specific point tasks for the air-delivery process based on the traditional air-to-surface fire control methods.Moreover, we construct the UAV maneuvering decision-making model based on Markov decision processes(MDPs). Specifically, we present a reward shaping method for the guidance towards area and guidance towards specific point tasks using potential-based function and expert-guided advice. The proposed algorithm could accelerate the convergence of the maneuvering decision-making policy and increase the stability of the policy in terms of the output during the later stage of training process. The effectiveness of the proposed maneuvering decision-making policy is illustrated by the curves of training parameters and extensive experimental results for testing the trained policy.

Decision-Making and Management of Self-Care in Persons with Traumatic Spinal Cord Injuries: A Preliminary Study

Patients and physicians understand the importance of self-care following spinal cord injury (SCI), yet many individuals with SCI do not adhere to recommended self-care activities despite logistical supports. Neurobehavioral determinants of SCI self-care behavior, such as impulsivity, are not widely studied, yet understanding them could inform efforts to improve SCI self-care. We explored associations between impulsivity and self-care in an observational study of 35 US adults age 18 - 50 who had traumatic SCI with paraplegia at least six months before assessment. The primary outcome measure was self-reported self-care. In LASSO regression models that included all neurobehavioral measures and demographics as predictors of self-care, dispositional measures of greater impulsivity (negative urgency, lack of premeditation, lack of perseverance), and reduced mindfulness were associated with reduced self-care. Outcome (magnitude) sensitivity, a latent decision-making parameter derived from computationally modeling successive choices in a gambling task, was also associated with self-care behavior. These results are preliminary;more research is needed to demonstrate the utility of these findings in clinical settings. Information about associations between impulsivity and poor self-care in people with SCI could guide the development of interventions to improve SCI self-care and help patients with elevated risks related to self-care and secondary health conditions.

Research on Public Engineering Emergency Decision-Making Based on Multi-Granularity Language Information

To effectively deal with fuzzy and uncertain information in public engineering emergencies,an emergency decision-making method based on multi-granularity language information is proposed.Firstly,decision makers select the appropriate language phrase set according to their own situation,give the preference information of the weight of each key indicator,and then transform the multi-granularity language information through consistency.On this basis,the sequential optimization technology of the approximately ideal scheme is introduced to obtain the weight coefficient of each key indicator.Subsequently,the weighted average operator is used to aggregate the preference information of each alternative scheme with the relative importance of decision-makers and the weight of key indicators in sequence,and the comprehensive evaluation value of each scheme is obtained to determine the optimal scheme.Lastly,the effectiveness and practicability of the method are verified by taking the earthwork collapse accident in the construction of a reservoir as an example.

A Blind Spot in the Reframing of a Universe of Possibles: Towards a Suitable Model for Decision-Making Theory and A.I.

Bayesian inference model is an optimal processing of incomplete information that, more than other models, better captures the way in which any decision-maker learns and updates his degree of rational beliefs about possible states of nature, in order to make a better judgment while taking new evidence into account. Such a scientific model proposed for the general theory of decision-making, like all others in general, whether in statistics, economics, operations research, A.I., data science or applied mathematics, regardless of whether they are time-dependent, have in common a theoretical basis that is axiomatized by relying on related concepts of a universe of possibles, especially the so-called universe (or the world), the state of nature (or the state of the world), when formulated explicitly. The issue of where to stand as an observer or a decision-maker to reframe such a universe of possibles together with a partition structure of knowledge (i.e. semantic formalisms), including a copy of itself as it was initially while generalizing it, is not addressed. Memory being the substratum, whether human or artificial, wherein everything stands, to date, even the theoretical possibility of such an operation of self-inclusion is prohibited by pure mathematics. We make this blind spot come to light through a counter-example (namely Archimedes’ Eureka experiment) and explore novel theoretical foundations, fitting better with a quantum form than with fuzzy modeling, to deal with more than a reference universe of possibles. This could open up a new path of investigation for the general theory of decision-making, as well as for Artificial Intelligence, often considered as the science of the imitation of human abilities, while being also the science of knowledge representation and the science of concept formation and reasoning.

Crossing the Achilles Heel of Algorithms:Identifying the Developmental Dilemma of Artificial Intelligence-Assisted Judicial Decision-Making

In the developmental dilemma of artificial intelligence(AI)-assisted judicial decision-making,the technical architecture of AI determines its inherent lack of transparency and interpretability,which is challenging to fundamentally improve.This can be considered a true challenge in the realm of AI-assisted judicial decision-making.By examining the court’s acceptance,integration,and trade-offs of AI technology embedded in the judicial field,the exploration of potential conflicts,interactions,and even mutual shaping between the two will not only reshape their conceptual connotations and intellectual boundaries but also strengthen the cognition and re-interpretation of the basic principles and core values of the judicial trial system.

Business Blockchain Suitability Determinants: Decision-Making through an Intuitionistic Fuzzy Method

Blockchain is one of the innovative and disruptive technologies that has a wide range of applications in multiple industries beyond cryptocurrency.The widespread adoption of blockchain technology in various industries has shown its potential to solve challenging business problems,as well as the possibility to create new business models which can increase a firm’s competitiveness.Due to the novelty of the technology,whereby many companies are still exploring potential use cases,and considering the complexity of blockchain technology,which may require huge changes to a company’s existing systems and processes,it is important for companies to carefully evaluate suitable use cases and determine if blockchain technology is the best solution for their specific needs.This research aims to provide an evaluation framework that determines the important dimensions of blockchain suitability assessment by identifying the key determinants of suitable use cases in a business context.In this paper,a novel approach that utilizes both qualitative(Delphi method)and quantitative(fuzzy set theory)methods has been proposed to objectively account for the uncertainty associated with data collection and the vagueness of subjective judgments.This work started by scanning available literature to identify major suitability dimensions and collected a range of criteria,indicators,and factors that had been previously identified for related purposes.Expert opinions were then gathered using a questionnaire to rank the importance and relevance of these elements to suitability decisions.Subsequently,the data were analyzed and we proceeded to integrate multi-criteria group decision-making(MCGDM)and intuitionistic fuzzy set(IFS)theory.The findings demonstrated a high level of agreement among experts,with the model being extremely sensitive to variances in expert assessments.Furthermore,the results helped to refine and select the most relevant suitability determinants under three important dimensions:functional suitability of the use case,organizational applicability,and ecosystem readiness.

分组密码算法uBlock积分攻击的改进

积分攻击是由Daemen等人(doi:10.1007/BFb0052343)于1997年提出的一种密码分析方法,是继差分分析和线性分析之后最有效的密码分析方法之一。作为2018年全国密码算法设计竞赛分组算法的获胜算法,uB-lock抵抗积分攻击的能力受到较多的关注。为了重新评估uBlock家族密码算法抵抗积分攻击的安全性,该文利用单项式传播技术,结合混合整数线性规划(MILP)工具搜索积分区分器,并利用部分和技术进行密钥恢复攻击。对于uBlock-128/128和uBlock-128/256,基于搜索到的9轮积分区分器分别进行了首个11轮和12轮攻击,数据复杂度为2~(127)选择明文,时间复杂度分别为2~(127.06)和2~(224)次加密,存储复杂度分别为2~(44.58)和2~(138)字节;对于uBlock-256/256,基于搜索到的10轮积分区分器进行了首个12轮攻击,数据复杂度为2~(253)选择明文,时间复杂度为2~(253.06)次加密,存储复杂度为2~(44.46)字节。与之前uBlock的最优积分攻击结果相比,uBlock-128/128和uBlock-256/256的攻击轮数分别提高2轮,uBlock-128/256的攻击轮数提高3轮。本文的攻击说明,uBlock针对积分攻击依然有足够的安全冗余。

Artificial intelligence promotes shared decision-making through recommending tests to febrile pediatric outpatients

BACKGROUND:To promote the shared decision-making(SDM)between patients and doctors in pediatric outpatient departments,this study was designed to validate artificial intelligence(AI)-initiated medical tests for children with fever.METHODS:We designed an AI model,named Xiaoyi,to suggest necessary tests for a febrile child before visiting a pediatric outpatient clinic.We calculated the sensitivity,specificity,and F1 score to evaluate the efficacy of Xiaoyi’s recommendations.The patients were divided into the rejection and acceptance groups.Then we analyzed the rejected examination items in order to obtain the corresponding reasons.RESULTS:We recruited a total of 11,867 children with fever who had used Xiaoyi in outpatient clinics.The recommended examinations given by Xiaoyi for 10,636(89.6%)patients were qualified.The average F1 score reached 0.94.A total of 58.4%of the patients accepted Xiaoyi’s suggestions(acceptance group),and 41.6%refused(rejection group).Imaging examinations were rejected by most patients(46.7%).The tests being time-consuming were rejected by 2,133 patients(43.2%),including rejecting pathogen studies in 1,347 patients(68.5%)and image studies in 732 patients(31.8%).The difficulty of sampling was the main reason for rejecting routine tests(41.9%).CONCLUSION:Our model has high accuracy and acceptability in recommending medical tests to febrile pediatric patients,and is worth promoting in facilitating SDM.

A multi-enterprise quality function deployment paradigm with unstructured decision-making in linguistic contexts

This paper presents an operational framework of unstructured decision-making approach involving quality function deployment(QFD)in an uncertain linguistic context.Firstly,QFD is extended to the multi-enterprise paradigm in a real-world manufacturing environment.Secondly,hesitant fuzzy linguistic term sets(HFLTSs),which facilitate the management and handling of information equivocality,are designed to construct a house of quality(HoQ)in the product planning process.The technique of computing with words is applied to bridge the gap between mechanisms of the human brain and machine processes with fuzzy linguistic term sets.Thirdly,a multi-enterprise QFD pattern is formulated as an unstructured decision-making problem for alternative infrastructure project selection in a manufacturing organization.The inter-relationships of cooperative partners are directly matched with a back propagation neural network(BPNN)to construct the multi-enterprise manufacturing network.The resilience of the manufacturing organization is considered by formulating an outranking method on the basis of HFLTSs to decide on infrastructure project alternatives.Finally,a real-world example,namely,the prototype manufacturing of an automatic transmission for a vehicle,is provided to illustrate the effectiveness of the proposed decision-making approach.

Compounding effects of human activities and climatic changes on coexistence of oasis-desert ecosystems:Prioritizing resilient decision-making for a riskier world

Water-salt balance is critical for the stable coexistence of salt-affected and groundwater-fed oasis-desert ecosystems. Yet, a comprehensive investigation of how soil salinization and groundwater degradation threaten the coexistence of oasis-desert ecosystems is still scarce, especially under the compounding effects of human activities and climatic changes. Here, we assessed the impacts of irrigated agriculture on hydrological regimes in oasisdesert systems, investigated the spatio-temporal variations of soil salinization in irrigated cropland, and evaluated the implications of the interplays of soil salinization and groundwater degradation on the coexistence of oasis-desert ecosystems in northwestern China, based on meaningful modelling approaches and comprehensive measurements over 1995–2020. The results showed that the irrigation return flow coefficient decreased sharply from 0.21 ± 0.09 in the traditional irrigation period to 0.09 ± 0.01 in the water-saving irrigation period. The continuous drop in groundwater tables and significant degradation of groundwater quality are occurring throughout this watershed. The eco-environmental flows are reaching to their limit with watershed closures(i.e.,the drainage from the oasis region into the desert region is being weakened or even eliminated), although these progressions were largely hidden by regional precipitation and streamflow variability. The process of salt migration and accumulation across different landscapes in oasis-desert system is being reshaped, and soil salinization in water-saving agricultural irrigated lands is accelerating with a regional average annual growth rate of18%. The vegetation in this watershed is degrading, and anthropogenic disturbance accelerates this trend. Our results highlight that environmental stress adaptation strategies must account for resilience maintenance to avoid accelerating catastrophic transitions in oasis-desert ecosystems. Determining the optimal oasis scales and formulating the best irrigation management plans are effective and resilient decision-making ways to maintain the coexistence relationship of oasis-desert ecosystem in drylands.

Twin-block在骨性Ⅱ类错[牙合]下颌后缩患儿中的应用

目的:探讨Twin-block在骨性Ⅱ类错[牙合]下颌后缩患儿中的应用效果。方法:选取2019年1月-2021年1月笔者医院接收的84例骨性Ⅱ类错[牙合]下颌后缩患儿设置为研究对象,按随机抽签法分为对照组、研究组,各42例,对照组采用直丝弓矫治器,研究组采用Twin-block矫治器。比较两组治疗前后咀嚼功能、牙齿咬合力、下颌骨变化、软组织面型变化、上气道矢状径,并统计家属治疗满意度。结果:治疗后,研究组咀嚼功能、牙齿咬合力、下颌骨变化指标、软组织面型变化指标均优于对照组(P<0.05);治疗后,研究组上气道矢状径高于对照组(P<0.05);研究组家属治疗满意度(95.24%)明显高于对照组(73.81%)(P<0.05)。结论:Twin-block应用于骨性Ⅱ类错[牙合]下颌后缩患儿治疗中,能改善咀嚼功能,增强牙齿咬合力,促进下颌骨生长,提高面型美观度,扩大上气道矢状径,利于改善呼吸功能,获得家属对治疗效果的高度评价。

Twin-Block双期矫治与单纯直丝弓矫正技术在早期下颌后缩矫正患儿中的疗效对比

目的探讨Twin-Block双期矫治与单纯直丝弓矫正技术在替牙及恒牙早期错(牙合)畸形矫正患儿中的疗效对比。方法选取我院2020年9月到2023年9月收治的60例恒牙早期错(牙合)畸形患儿作为研究对象,按照治疗方法不同分为观察组与对照组,各组均为30例。所有患儿均采取矫正治疗,对照组患儿采取直丝弓矫正技术进行治疗,观察组患儿采取Twin-Block矫治器及直丝弓矫治技术进行双期治疗。对比两组患儿临床疗效,及矫治时间,治疗前头影测量及关节改建情况,采用口腔锥形束扫描成像进行检测两组患儿关节窝宽径、关节前间隙、关节后间隙、关节上间隙、髁突前后径、髁突高度相关参数表达水平,最后对比两组患儿治疗前后上下齿槽座(ANB)角、下齿槽座(SNB)角、上齿槽座(SNA)角、下颌平面角(Mp-SN)、下颌综合长度(Ar-Po)、下颌升支高度(Ar-Go)、下颌体长(Go-Gn)、上中切牙-SN平面角(U1-SN)、下中切牙-下颌平面角(L1-Mp)、上唇审美平面距(Ls-E)、下唇审美平面距(Li-E)、鼻唇角。结果观察组治疗总有效率96.67%明显高于对照组76.67%(P<0.05);治疗后,观察组矫治时间长于对照组(P<0.05);治疗前,两组患儿关节窝宽径、关节前间隙、关节后间隙、关节上间隙、髁突前后径、髁突高度对比无明显差异(P>0.05),治疗后,两组患儿关节窝宽径对比无明显差异,关节前间隙减小,观察组(1.21±0.28)mm高于对照组,关节后间隙、关节上间隙、髁突前后径、髁突高度均增加,且观察组高于对照组(P<0.05);治疗前,两组患儿ANB角、SNB角、SNA角、Mp-SN、Ar-Po、Ar-Go、Go-Gn、U1-SN、L1-Mp、Ls-E、Li-E、鼻唇角对比无明显差异(P>0.05),治疗后,两组患儿SNB、Mp-SN、Ar-Po、Ar-Go、Go-Gn、鼻唇角均增大,ANB、U1-SN、L1-Mp、Li-E、Ls-E减小,且观察组变化大于对照组(P<0.05)。结论Twin-Block矫治器配合直丝弓矫治技术双期矫治器与单纯直丝弓矫正技术相比治疗替牙及恒牙早期错(牙合)畸形疗效显著,可作为儿童恒牙早期错(牙合)畸形的首选治疗方案。

基于Group-Res2Block的智能合成语音说话人确认方法

针对现有说话人确认任务基于自然语音条件下并不适用于智能合成语音的问题,提出一种基于Group-Res2Block的智能合成语音说话人确认方法。首先,设计了Group-Res2Block结构,在Res2Block的基础上将当前分组与相邻前后分组进行合并形成新的分组,以增强说话人局部特征的上下文联系;其次,设计了并行结构的多尺度通道注意力特征融合机制,利用不同大小卷积核实现同一层级的特征在通道维度的特征选择,以获取更具表现力的说话人特征,避免信息冗余;最后,设计了串行结构的多尺度层注意力特征融合机制,构建层结构,将深浅层特征整体进行融合并赋予不同权重,以获取最优的特征表达。为验证所提出特征提取网络的有效性,构建了中英文两种智能合成语音数据集进行消融实验和对比实验。结果表明本文方法在该任务的评价指标精确度(accuracy,ACC)、等错误率(equal error rate,EER)和最小检测代价函数(minimum detection cost function,minDCF)上是最优的。此外,通过对模型泛化性能进行测试,验证了本文方法对未知智能语音算法的适用性。