基于用户关注空间与注意力分析的视频精彩摘要与排序

文中提出一种基于用户关注空间与注意力分析的视频内容理解方法,该方法可以有效地获得多通道的视频关注信息,并可使用户根据个性化需求定制视频关注内容,实现视频的高效浏览与访问.首先采用基于二叉层次型结构与分类器选择的音频分类算法将视频中的主要声音类型分类,然后将视频中影响用户注意力的视觉、听觉、时序因素定义为用户关注空间,分别使用相应的中层特征在这三个方面对用户注意力进行表示并计算其关注度,从而在音视频底层特征与高层认知之间建立有机过渡.作者设计了顺序决策融合算法来融合视觉与听觉关注度,生成关注度时序变化曲线并获得精彩摘要.最后使用支持向量回归模型并引入相关反馈机制来实现用户个性化的精彩片段排序.该项工作的特点是通过建立符合人类认知规律的关注度模型并结合相关反馈技术,对视频内容进行类人理解.实验证明,该方法对提取与生成符合用户个性化要求的视频摘要及排序结果具有良好的效果.     

个性化服务用户模型研究

以数字图书馆为研究对象,提出了一种个性化服务用户模型构架,并对实现过程中的几表示方法、用户模型的建立以及更新算法进行了详细论述,最后在个性化文本过滤算法基础上,得到在实际的数字图书馆中的验证结果.用户兴趣的提取采用支持向量机分类算法和无监督聚类算法相结合的隐式方式获得;在考虑最近到达的兴趣与用户原有兴趣序列的综合影响的基础上,用户兴趣的更新采用最近最少使用淘汰算法.实验结果表明,该模型具有隐式荻取用户兴趣、用户模型更新命中率高等特点.     

融合评分矩阵和评论文本深度神经网络推荐模型

针对个性化推荐中用户评分矩阵数据集稀疏,用户和项目描述信息未充分利用的问题,提出融合评分矩阵和评论文本的深度神经网络推荐模型(deep neural network recommendation model,DeepRec)。首先将通过数据预处理得到的用户偏好特征和项目属性特征的文本集合分别输入到卷积神经网络进行训练,得到用户和项目的深层次非线性特征,同时将评分矩阵输入多层感知机得到用户偏好隐表示,并对两种模型提取的用户偏好隐表示进行融合;其次利用多层感知机建模用户和项目隐表示对用户进行个性化推荐;最后基于三组数据集以均方根误差为评估指标进行对比实验。结果表明DeepRec的预测误差更低,有效提高了推荐的精准度。     

个性化服务中网页推荐模型的研究

提出了一种新颖的个性化网页推荐模型。它融合了信息检索、信息过滤和数据挖掘中的相关技术。运用两种不同的特征向量来表示用户兴趣,并且通过聚类生成群体兴趣扩展了个体兴趣。同时阐述了用户兴趣文件建立、维护和个性化网页推荐。实验证明了推荐模型的准确性和有效性。     

基于用户特定空间的对抗推荐模型

自编码器和生成对抗网络作为强大的模型已经被应用到推荐系统领域,能补充用户项目之间的交互信息。但这种模式的训练下,大量的辅助信息被浪费,比如用户特定信息。结合自编码器、生成对抗网络和辅助信息,提出了基于用户特定空间的对抗推荐模型。为建立辅助信息和交互信息的联系,将自编码器的隐空间替换为用户特定空间,交互信息和用户特定信息的点对点映射又限制了模型的表达,因此,在用户特定空间加入对抗训练增加模型的性能。在两个公开电影数据集的充分实验证明了提出模型的有效性和优越性。     

基于短期记忆与遗忘系数的用户个性化建模方法研究

搜索引擎的一个标准是不同的用户用相同的查询条件检索时,返回的结果相同。为解决准确性问题,个性化搜索引擎被提出,它可以根据用户的不同个性化特征提供不同的搜索结果。然而,现有的方法更注重用户的长时记忆和独立的用户日志文件,从而降低了个性化搜索的有效性。获取用户短时记忆模型来提供准确有效的用户偏好的个性化搜索方法被广泛采用。首先,根据基于查询关键词的相关概念生成短期记忆模型;接着,基于用户的时序有效点击数据生成用户个性化模型;最后,在用户会话中引入了遗忘因子来优化用户个性化模型。实验结果表明,所提出的方法可以较好地表达用户信息需求,较为准确地构建用户的个性化模型。     

基于模糊兴趣模型与多Agent的个性化推荐系统

提出了基于模糊兴趣模型与多Agent的个性化推荐系统框架,通过引入用户模糊兴趣模型,使以Agent为基础的推荐系统无法通过隐式收集用户对商品属性评价的问题得到解决,并且在客户端收集并挖掘用户的私有信息,然后从服务器中获取用户感兴趣的信息,最后生成并更新UserProfile。     

关联规则算法在Web个性化服务中的研究与应用

重点研究了Web日志挖掘,提出了一个Web个性化信息挖掘模型,设计了某高校图书馆个性化服务系统My Library。系统采用关联规则挖掘算法,从服务器日志中得到用户感兴趣的隐式模式,并将该隐式兴趣集推荐给用户,从而在一定程度上实现了个性化服务。     

面向移动应用的探索式个性化服务推荐方法

个性化推荐系统是应用系统中广泛应用的技术之一,用户兴趣偏好模型的建立与更新是个性化推荐系统的关键环节,针对移动设备位置随时变化的特点,以移动端的应用系统为研究对象,提出了一种随用户位置变化而动态更新的用户兴趣偏好模型,并对实现过程中的几个关键问题,包括用户兴趣偏好模型表示方法、用户兴趣关键字提取、用户兴趣偏好模型的建立与更新算法进行了详细描述,最后利用用户兴趣偏好模型根据协同过滤算法进行个性化推荐,并根据用户对推荐结果的评价进一步修正用户兴趣偏好模型.用户兴趣偏好模型采用基于兴趣关键字的向量空间模型表示,用户兴趣关键字由根据TF-IDF算法获得的用户隐式兴趣和用户参与的显式兴趣相结合获得,用户位置信息变化时,系统获取当前位置附近的服务,对已存在于用户兴趣关键树中的服务权值进行增强,而对不存在其中的进行遗忘以调整用户兴趣树从而更新用户兴趣偏好模型.验证表明,该方法推荐的服务更符合用户所处的位置上下文环境,并且具有高度的可达性.     

个性化搜索引擎中的用户兴趣模型分析与研究

针对用户兴趣模型构建问题,利用用户兴趣树描述用户兴趣,采用空间向量模型的表示方法,对构建用户个性化模型进行研究,并提出一种兴趣模型调整算法。模拟实验表明,该模型能有效提高检索结果的查准率,以满足用户个性化需求。     

The effect of head-related transfer function measurement methodology on localization performance in spatial audio interfaces

OBJECTIVE: Four head-related transfer function (HRTF) data sets were compared to determine the effect of HRTF measurement methodology on the localization of spatialized auditory stimuli. BACKGROUND: Spatial audio interfaces typically require HRTF data sets to generate the spatialized auditory stimuli. HRTF measurement is accomplished using a variety of techniques that can require several nearly arbitrary decisions about methodology. The effects of these choices upon the resulting spatial audio interface are unclear. METHOD: Sixteen participants completed a sound localization task that included real-world, broadband stimuli spatialized at eight locations on the horizontal plane. Four different HRTF data sets were utilized to spatialize the stimuli: two publicly available HRTF data sets and two data sets obtained using different in-house measurement systems. All HRTFs were obtained from the Knowles Electronics Mannequin for Acoustics Research. RESULTS: Unsigned localization error and proportion of front/back reversals did not differ significantly across HRTF data sets. Poorest accuracy was observed in locations near the medial (front/back) axis of the listener, mainly because of the relatively large proportions of reversals at these locations. CONCLUSION: This study suggests that the particular generalized HRTF data set chosen for spatialization is of minimal importance to the localizability of the resulting stimuli. APPLICATION: This result will inform the design of many spatial audio interfaces that are based upon generalized HRTFs, including wayfaring devices, communication systems, and virtual reality systems.     

基于小波神经网络的HRTFs逼近与声频仿真

头部相关传递函数是指从声源到耳鼓的谱滤波器,它因提取了声源的方位信息,所以在声频仿真中,它是非常重要的立体听觉定位曲线。由于ＨＲＴＦｓ随声源的相对位置、频和听觉对象不同而变化,并与其自变量之间还存在着非常复杂的非线性关系,所以三维声音仿真的实现需要处理庞大的ＨＲＴＦｓ的数据。     

基于可解释网络解耦表征的低成本雷达定位解算方法

为降低调频连续波（FMCW）雷达成本,同时提高定位精度,设计可解释解耦表征模型,该模型由网络解算器、虚假信号生成器以及可解释潜变量三部分组成,首先处理雷达信号获得中频频谱;然后输入到网络解算器中生成位置潜变量;再通过物理机制对潜变量进行转换,生成虚假中频信号频谱;最后,设计局部光滑损失函数对模型进行自监督训练,实现潜变量的解耦物理表征。实验结果表明：所提算法能对雷达系统频谱信号的粗粒度进行超分辨率细化,其机理能有效应对雷达系统的硬件公差、环境噪声、安装误差等问题,并可自动地训练出雷达的解算网络,从而具有大规模室内、机载联网定位的应用潜力。     

Anthropometric-based clustering of pinnae and its application in personalizing HRTFs

A huge amount of anthropometric data on pinnae have been widely used in ergonomic designs, but their usage for personalizing head-related transfer functions (HRTFs) to improve virtual auditory display has received insufficient attention. The present work proposes a simplified method of HRTFs selection based on pinnae clustering. It uses a large amount of pinna anthropometric data but only HRTFs of several typical subjects with cluster-center pinnae. A baseline database with pinnae of 100 subjects was clustered into eight groups, where eight typical subjects were identified as the cluster centers. According to the similarity of the pinnae anthropometric parameters of a new listener and eight typical subjects, the HRTFs of the best-matched subject were selected as the pair of personalized HRTFs for the new listener. The errors in the magnitude, peaks, and notches of the HRTF spectra show that the matched HRTFs are closer to the new subject's own HRTFs, compared with the HRTFs of other cluster-center subjects. The subjects have a low in-head localization rate and confusion rate using matched HRTFs in the psychoacoustic experiment. This method decreases the workload since we need to acquire fewer HRTF data and it is possible to use a huge amount of pinna data in customizing HRTFs. Our results provide evidence for potential applications in ergonomic design.     

Rendering localized spatial audio in a virtual auditory space

High-quality virtual audio scene rendering is required for emerging virtual and augmented reality applications, perceptual user interfaces, and sonification of data. We describe algorithms for creation of virtual auditory spaces by rendering cues that arise from anatomical scattering, environmental scattering, and dynamical effects. We use a novel way of personalizing the head related transfer functions (HRTFs) from a database, based on anatomical measurements. Details of algorithms for HRTF interpolation, room impulse response creation, HRTF selection from a database, and audio scene presentation are presented. Our system runs in real time on an office PC without specialized DSP hardware.     

声学仿真中的人工神经网络方法

虽然许多研究人员已认识到三维真实感声音在未来人机交互中的重要地位，但是三维真实感声音在计算机领域的真正实现仍有不少障碍有待克服．基于对声学及心理声学最新研究成果的调查和分析，本文设计并实现了一个基于神经网络方法的ＨＲＴＦ（ｈｅａｄ－ｒｅｌａｔｅｄｔｒａｎｓｆｅｒｆｕｎｃｔｉｏｎ）模型，用于三维真实感声音的生成．模型中的数据可通过网络学习进行重新设置，以满足多种场合的需要．并且，通过神经网络的非线性拟合能力，可以获取空间任意位置的ＨＲＴＦ数据．初步的实验表明了该方法的有效性和正确性．     

Effect of Emulated Head-Tracking for Reducing Localization Errors in Virtual Audio Simulation

Virtual audio simulation uses head-related transfer function (HRTF) synthesis and headphone playback to create a sound field similar to real-life environments. Localization performance is influenced by parameters such as the recording method and the spatial resolution of the HRTFs, equalization of the measurement chain as well as common headphone playback errors. The most important errors are in-the-head localization and front-back reversals. Among other cues, small movements of the head are considered to be important to avoid these phenomena. This study uses the BEACHTRON sound card and its HRTFs for emulating small head-movements by randomly moving the virtual sound source to emulate head-movements. This method does not need any additional equipment, sensors, or feedback. Fifty untrained subjects participated in the listening tests using different stimuli and presentation speed. A virtual target source was rendered in front of the listener by random movements of 1c-7deg. Experiments showed that this kind of simulation can be helpful to resolve in-the-head localization, but there is no clear benefit for resolving front-back errors. Emulation of small head-movements of 2deg could actually increase externalization rates in about 21% of the subjects while presentation speed is not significant.     

与头相关联的传递函数的测量及数据分析

关于空间听觉的研究表明对“与头相关联的传递函数（ＨＲＴＦ）”进行测量和研究备受关注。在自由声场声源的条件下，测量是由左右耳耳膜处的记录信号组成，测得的ＨＲＴＦ幅度特性的变化为声源位置的函数。本文对近年来的一些测量方法和测量数据作了研究分析     

基于中国人工头模的HRTF数据库设计与实现

头相关传输函数(Head Related Transfer Function,HRTF)描述了在自由场情况下,点声源到人耳鼓膜处的传输过程,其中包含有重要的声源定位信息。本文搭建HRTF测量与实验环境,设计和实现一个HRTF数据库,数据库包含中国人平均头模BHead210 481个空间方位的头相关脉冲响应(Head Related Impulse Response)数据。进行主观定位判听实验,比较BHead210人工头测量的HRTF数据和KEMAR人工头HRTF数据在中国受试者上的判听效果。     

State-Space Synthesis of Virtual Auditory Space

Binaural displays for immersive listening must model realistic acoustic environments, multiple sound sources, and accommodate source and head motion. Many displays accomplish this by convolving collections of spatially distributed point sources with head-related transfer functions (HRTFs). The computational load of such a system scales linearly with the number of HRTFs modeled by the display. Realistic scenes often require a large number of HRTFs, and this framework is computationally burdensome. We propose a method that significantly eases this load by formulating the HRTF filter array as a state-space system. Three state-space architectures are explored. The performance of the most general architecture is found to suffer due to the interaural time delay (ITD). This problem may be circumvented with two alternative architectures, although the ideal choice depends on the specific display application. For each architecture, two order-reduction techniques are explored. Both techniques are based on the Hankel operator; one is ad hoc and simple to implement whereas the other is optimal in the Hankel norm. The two methods yield similar auditory performance, although the optimal method may be desirable for HRTF approximation.