随着移动通信技术的发展演进,6G(6th-Generation)网络作为新一代智能化数字信息基础设施,将不再仅聚焦信号的传输和复现,更需要基于电磁传播过程实现对周围环境的高效感知和理解,从而获取信道语义知识,协助智能通信体的预测、决策、波束成形等.因此,相较于传统信道而言,赋予无线信道模型对物理环境的语义理解、重构、表达能力,已成为智能无线信道模型的重要特征.本文提出了一种无线信道语义的分析和建模方法,将信道语义定义为状态语义、行为语义和事件语义3种层级,分别对应信道瞬态多径、信道时变轨迹和信道拓扑结构.此外,基于车载通感一体化(Integrated Sensing And Communication,ISAC)信道测量系统,开展了28 GHz下面向信道语义表征的无线信道测量,基于实测数据对信道语义进行解构、标识、建模,重点分析了3种不同语义下的信道多径分布特性,完成了语义导向的信道生成,结果表明信道语义模型能够在生成较准确信道的同时,表达更丰富的语义信息.本文工作是在语义层面上探索智能信道建模的新方法,通过深入挖掘无线信道的内在语义特征,促进通信系统在理解和认知环境方面的能力,从而提高通信效率和质量.
为解决可逆信息隐藏(reversible data hiding,RDH)容量受限的问题,提出了一种基于相邻均值差的可逆信息隐藏(neighboring mean difference reversible data hiding,NMDRDH)算法。相邻均值差(neighboring mean diffe-rence,NMD):计算两...
详细信息
为解决可逆信息隐藏(reversible data hiding,RDH)容量受限的问题,提出了一种基于相邻均值差的可逆信息隐藏(neighboring mean difference reversible data hiding,NMDRDH)算法。相邻均值差(neighboring mean diffe-rence,NMD):计算两个相邻数值的平均值与其中一个数值的差值。NMD将差值最小化,使数据更加集中。首先将图像进行分块,然后在分块上应用NMD生成差值直方图,最后通过平移差值直方图,利用峰值点来嵌入秘密信息。由于NMD使生成的差值直方图具有更多的峰值点,所以该方法可嵌入更多的秘密信息。实验结果表明,采用本算法,原始图像恢复率和秘密信息提取正确率均为100%;相比于经典差值直方图平移方法,本算法的嵌入容量提升了43.7%;本算法在保证高容量的同时,PSNR达到42 dB以上,确保了嵌入图像失真较小。
随着人工智能技术的迅猛发展,医疗问答系统已成为医疗信息检索和知识获取的重要工具。医疗领域涉及大量医学术语、复杂的疾病症状和治疗方案,传统查询方式难以高效、准确地满足医护人员和患者的信息需求。相比传统国内搜索引擎和原生开源大语言模型(LLMs),基于LangChain的大模型医疗问答系统能够提供更高质量的答案,显著提升医疗知识检索的效率和精准度。因此,本研究提出了一种基于LangChain与大模型的医疗智能问答系统,结合命名实体识别(NER)、图谱查询和对话分析等技术,构建了一个专注于医疗领域的知识图谱及其查询与生成模块。通过设计和优化Prompt提示词,Agent Tool提升了大模型生成更精准、高质量医疗问答的能力。研究结果表明,该系统在医疗问答任务中的表现优异,准确度、方案可行性和上下文相关性等指标显著优于传统LLMs和国内知名大模型。该系统通过与大规模医疗知识图谱的结合,能够深入理解复杂的医疗问题,并提供精准的回答,呈现可视化图谱展示图,更直观地给用户反馈,同时具备较高的数据安全性和可迁移性。Nowadays, with the rapid development of artificial intelligence technology, medical question answering system has become an important tool for medical information retrieval and knowledge acquisition. The medical field involves a large number of medical terms, complicated disease symptoms and treatment plans, and traditional inquiry methods are difficult to meet the information needs of medical staff and patients efficiently and accurately. Compared with traditional domestic search engines and native open source large language model (LLMs), LangChain-based large model medical question answering system can provide higher quality answers, significantly improving the efficiency and accuracy of medical knowledge retrieval. Therefore, this study proposed a medical intelligent question and answer system based on LangChain and large model, combined with named entity recognition (NER), graph query and dialogue analysis and other technologies, to build a knowledge graph and query and generation module focusing on the medical field. By designing and optimizing Prompt words, Agent Tool improves the ability of large models to generate more accurate and high-quality medical questions and answers. The results show that the system performs well in medical question answering tasks, with significant improvements in accuracy, feasibility, and context relevance are significantly better than traditional LLMs and well-known domestic large models. Through the combination of large-scale medical knowledge graph, the system can deeply understan
暂无评论