应用改进混合蛙跳算法的实用语音情感识别

针对支持向量机(Support Vector Machine,SVM)的参数优化问题,提出了一种改进的混合蛙跳算法(Improved Shuffled Frog Leaping Algorithm,Im-SFLA),提高了其在实用语音情感识别中的学习能力。首先,我们在SFLA中引入了模拟退火(Simulated Annealing,SA)、免疫接种(Immune Vaccination,IV)、高斯变异和混沌扰动算子,平衡了搜索的高效性和种群的多样性;第二,利用Im-SFLA优化SVM的参数,提出了一种Im-SFLA-SVM方法;第三,分析了烦躁等实用语音情感的声学特征,重点分析了基音、短时能量、共振峰和混沌特征随情感类别的变化特性,构建出144维的情感特征向量并采用LDA降维到4维;最后,在实用语音情感数据库上测试了算法性能,将提出的算法与混合蛙跳算法(Shuffled Frog Leaping Algorithm,SFLA)优化SVM参数的方法(SFLA-SVM方法)、粒子群优化(Particle Swarm Optimization,PSO)算法优化SVM参数的方法(PSO-SVM方法)、基本SVM方法、高斯混合模型(Gaussian Mixture Model,GMM)方法和反向传播(Back Propagation,BP)神经网络法等进行对比。实验结果表明,采用Im-SFLA-SVM方法的平均识别率达到77.8%,分别高于SFLA-SVM方法、PSO-SVM方法、SVM方法、GMM方法和BP神经网络法各1.7%,2.7%,3.4%,4.7%,7.8%,并且对于烦躁这种实用情感的识别率提高效果最为明显,从而证实了Im-SFLA是一种有效的SVM参数选择方法,并且Im-SFLA-SVM方法能显著提升实用语音情感的识别率。      

3维电磁粒子模拟程序设计

 在3维Yee网格模型以及蛙跳模型的基础上，分析了3维电磁场离散算法及其稳定条件，讨论了3维空间网格上电磁场作用于粒子的平均场权重法和粒子作用于电磁场的Charge-conserving法，同时也讨论了3维空间网格上理想导体边界条件、自由吸收边界条件、爆炸式发射模型的模拟实现方法。以向内发射同轴二极管为例从其电流电压关系及电子束运动特性方面验证了模拟的正确性。     

非线性共振及束晕-混沌的场强参数特性

通过计算和分析周期聚焦磁场强度和表征束自生场强度的束导流系数的变化对强流粒子束运动特征的影响,系统研究了强流粒子束的非线性共振和束晕现象。采用庞加莱截面技术对束核包络振荡及其单粒子运动轨迹进行了数值模拟,清楚地展示了不同情况下束核包络非线性振荡以及对应的试验粒子空间分布。结果表明:束核自生场使得束核包络发生非线性振荡,随着束核自生场和聚焦磁场的增加,伴随着束核共振岛的出现,单粒子空间分布出现越来越多的束晕粒子。     

基于粒子滤波的单通道正弦调频混合信号分离与参数估计

以单通道正弦调频(SFM)混合信号为研究对象,提出了基于粒子滤波的正弦调频混合信号分离与参数提取方法.针对正弦调频混合信号频率无跳变的特征,提出了一种基于粒子滤波的相位差解混叠算法,并通过源信号相位差解决了本算法中粒子滤波高维状态空间降维问题,提出了一种适合高维状态空间的似然函数模型,比较固定长度粒子估计值和真实值误差,进而准确衡量粒子权重.通过在重采样后引入MCMC转移,解决了静止参数下粒子多样性降低问题,有效提高粒子滤波迭代收敛速度.从而在先验知识仅已知信号调制方式的情况下,完成对单通道正弦调频混合信号的参数提取,并通过重构信号完成正弦调频混合信号分离.最后通过仿真分析发现,该方法能够有效的实现正弦调频混合信号的分离与参数估计.     

基于STA-CRNN模型的语声情感识别*

语声情感识别对人机交互和情感计算研究领域具有重要作用,各类研究方法层出不穷。近期研究学者应用卷积神经网络和长短期记忆网络方法提取对数Mel谱图空间特征和时间特征,取得了一定的成果。然而不论是卷积神经网络还是长短期记忆网络提取特征时,都会产生特征冗余,导致语声情感识别效果下降。针对这一问题,该文提出了一种基于时空注意力机制的卷积-递归神经网络模型,采用对数Mel谱图和其一阶差分、二阶差分作为特征输入,在使用卷积神经网络提取空间特征和长短期记忆网络提取时间特征时,加入空间注意力和时间注意力机制,从而使上述网络能够更好地提取到对数Mel谱图中有效表征情感的空间特征和时间特征。该模型在Emo-DB和IEMOCAP语声数据集上的加权准确率分别达到86.8%、69.4%,未加权准确率分别达到84.7%、65.5%,优于当前大多数先进方法。     

居中热库蛙跳算法的大尺度模拟测试

分子动力学模拟一个必不可少的要素是对运动方程的准确高效积分. 尽管在动力学传播框架内的现代计算工具中大量使用了传统的蛙跳算法,研究表明蛙跳算法仍然可以进一步改进以获得更好的性能. 居中控温方案中蛙跳的改进版本(蛙跳-middle)实现了更高的精度和效率,即使积分时间步长扩大了几倍,也能保持稳定的动力学传播. 本文使用常规和增强采样模拟中对两个积分算法(蛙跳和蛙跳-middle)进行大尺度模拟测试,旨在标定由时间步长引起的全局性质(例如,详细的势能项)的变化行为以及复杂系统的实际模拟中的局部可观察量(例如,自由能变化或键长). 测试集由六个化学和生物相关系统组成,包括$N$-甲基乙酰胺和A₇-T₇DNA中二面角翻转的构象变化、丙二醛内的分子内质子转移、苯和苯酚针对T4溶菌酶L99A的结合自由能计算、乙胺醇低共熔溶剂中的羟基键长变化以及蓝光受体蛋白的势能计算. 所有结果显示蛙跳-middle积分算法中时间步长引起的误差较小. 蛙跳-middle相对传统蛙跳积分算法的性能提高对于全局属性比局部可观测量更大. 总体而言,目前的工作结果表明,在实际化学和生物系统的模拟中,应优先应用蛙跳-middle算法以获得准确的热力学行为和性质.     

碎片超高速撞击防护结构粒子场三维重构

为模拟空间碎片超高速撞击航天器防护结构表面材料喷射/溅射粒子场演化过程,并获取粒子场相关物理信息,基于粒子场同轴激光全息图像开展了碎片撞击过程的三维重构技术研究.首先对全息图像进行边缘剪切和缩放,将其划分为分辨率300×300左右的子图像以便于进行网格剖分;对于粒子堆叠区域子图像,采用基于三角化的网格剖分算法;剖分后形成的单一粒子采用Sobel算子提取其二维轮廓,然后将其投影到特定的三维空间形成三维形体;基于MAXScript语言实现了粒子场演化过程模拟.重构结果表明,无论是粒子场静态三维重构结果还是其演化过程均与撞击试验全息图像吻合较好,从而验证了该重构技术的有效性,为研究空间碎片对航天器防护结构的损伤效应提供了一种新的思路.     

一种基于离散粒子群优化算法的高光谱图像端元提取方法

针对混合像元分解过程中,由于数据噪声引起的端元提取不准确问题,引入了群智能算法中的粒子群优化算法,并对粒子群优化算法进行了改进,重新定义了位置和速度的表示方法和更新策略,得到离散粒子群优化(discrete particle swarm optimization,D-PSO),能够在离散空间中进行搜索,解决组合优化问题。同时,通过定义目标函数和可行解空间,将端元提取问题改写成组合优化问题,最终实现利用D-PSO进行端元提取。在给出算法的详细流程之后,文章通过一组模拟数据实验和一组实际数据实验验证了D-PSO算法对于具有较大噪声的数据的适应性和提取端元的可信程度,并分析了不同参数对于算法性能的影响。     

碎片超高速撞击防护结构粒子场三维重构

为模拟空间碎片超高速撞击航天器防护结构表面材料喷射/溅射粒子场演化过程,并获取粒子场相关物理信息,基于粒子场同轴激光全息图像开展了碎片撞击过程的三维重构技术研究.首先对全息图像进行边缘剪切和缩放,将其划分为分辨率300×300左右的子图像以便于进行网格剖分;对于粒子堆叠区域子图像,采用基于三角化的网格剖分算法;剖分后形成的单一粒子采用Sobel算子提取其二维轮廓,然后将其投影到特定的三维空间形成三维形体;基于MAXScript语言实现了粒子场演化过程模拟.重构结果表明,无论是粒子场静态三维重构结果还是其演化过程均与撞击试验全息图像吻合较好,从而验证了该重构技术的有效性,为研究空间碎片对航天器防护结构的损伤效应提供了一种新的思路.     

基于混合粒子滤波的水下小目标跟踪

针对水下小目标粒子滤波估计过程中“粒子贫化”引起的估计性能下降,提出了混合粒子滤波算法。该算法在常规粒子滤波算法基础上,在每一步迭代估计过程中进行量测的再次随机采样,以丰富随机粒子多样性,缓解水下小目标状态估计过程中的“粒子贫化”的影响。对算法进行了仿真分析,并将该方法用于水下小目标探测实验的数据处理,结果表明,相比于常规的粒子滤波算法,所提出的混合粒子滤波得到了误差更小且稳定的状态估计结果,有效地改善水下小目标跟踪的精度和稳健性。     

A simpler and elegant algorithm for computing fractal dimension in higher dimensional state space

Chaotic systems are now frequently encountered in almost all branches of sciences. Dimension of such systems provides an important measure for easy characterization of dynamics of the systems. Conventional algorithms for computing dimension of such systems in higher dimensional state space face an unavoidable problem of enormous storage requirement. Here we present an algorithm, which uses a simple but very powerful technique and faces no problem in computing dimension in higher dimensional state space. The unique indexing technique of hypercubes, used in this algorithm, provides a clever means to drastically reduce the requirement of storage. It is shown that theoretically this algorithm faces no problem in computing capacity dimension in any dimension of the embedding state space as far as the actual dimension of the attractor is finite. Unlike the existing algorithms, memory requirement offered by this algorithm depends only on the actual dimension of the attractor and has no explicit dependence on the number of data points considered.     

Neural network modeling of emotion

This article reviews the history and development of computational neural network modeling of cognitive and behavioral processes that involve emotion. The exposition starts with models of classical conditioning dating from the early 1970s. Then it proceeds toward models of interactions between emotion and attention. Then models of emotional influences on decision making are reviewed, including some speculative (not and not yet simulated) models of the evolution of decision rules. Through the late 1980s, the neural networks developed to model emotional processes were mainly embodiments of significant functional principles motivated by psychological data. In the last two decades, network models of these processes have become much more detailed in their incorporation of known physiological properties of specific brain regions, while preserving many of the psychological principles from the earlier models.Most network models of emotional processes so far have dealt with positive and negative emotion in general, rather than specific emotions such as fear, joy, sadness, and anger. But a later section of this article reviews a few models relevant to specific emotions: one family of models of auditory fear conditioning in rats, and one model of induced pleasure enhancing creativity in humans. Then models of emotional disorders are reviewed. The article concludes with philosophical statements about the essential contributions of emotion to intelligent behavior and the importance of quantitative theories and models to the interdisciplinary enterprise of understanding the interactions of emotion, cognition, and behavior.     

一种音乐情绪参数化的方法

针对目前基于情绪的音乐分类研究存在的弊端,为了方便音乐检索,本文提出一种音乐情绪参数化的方法。该方法通过提取反映音乐情绪的特征向量,然后利用fisher算法进行维数压缩,再通过大量的音乐样本训练得到节奏、音调和音色3个描述音乐情绪的参数,参数的大小反映了情绪的强弱。实验结果表明,音乐情绪参数化的结果符合音乐实际的情绪。     

面向语音情感识别的改进可辨别完全局部二值模式

为了研究语音情感与语谱图特征间的关系,本文研究并提出一种面向语音情感识别的改进可辨别完全局部二值模式特征。首先,基于语谱图灰度图像,计算图像的完全局部二值符号模式(CLBP_S)、幅度模式(CLBP_M)的统计直方图。然后,将CLBP_S,CLBP_M统计直方图输入可区别特征学习模型中,训练得到全局显著性模式集合。最后,采用全局显著性模式集合对CLBP_S,CLBP_M直方图进行处理,将处理后的特征级联,得到面向语音情感识别的改进可辨别完全局部二值模式特征(IDisCLBP_SER)。基于柏林库、中文情感语音库的语音情感识别实验显示,IDisCLBP_SER特征召回率比纹理图像信息(TII)等特征提高了8%以上,比声学频谱特征平均提高了4%以上。而且,本文提出的特征可以和现有声学特征进行较好融合,融合后的特征召回率比现有声学特征召回率提高1%~4%。      

Cryptographic Algorithm Using Newton-Raphson Method and General Bischi-Naimzadah Duopoly System

Image encryption is an excellent method for the protection of image content. Most authors used the permutation-substitution model to encrypt/decrypt the image. Chaos-based image encryption methods are used in this model to shuffle the rows/columns and change the pixel values. In parallel, authors proposed permutation using non-chaotic methods and have displayed good results in comparison to chaos-based methods. In the current article, a new image encryption algorithm is designed using combination of Newton-Raphson’s method (non-chaotic) and general Bischi-Naimzadah duopoly system as a hyperchaotic two-dimensional map. The plain image is first shuffled by using Newton-Raphson’s method. Next, a secret matrix with the same size of the plain image is created using general Bischi-Naimzadah duopoly system. Finally, the XOR between the secret matrix and the shuffled image is calculated and then the cipher image is obtained. Several security experiments are executed to measure the efficiency of the proposed algorithm, such as key space analysis, correlation coefficients analysis, histogram analysis, entropy analysis, differential attacks analysis, key sensitivity analysis, robustness analysis, chosen plaintext attack analysis, computational analysis, and NIST statistical Tests. Compared to many recent algorithms, the proposed algorithm has good security efficiency.     

Explicit attention interferes with selective emotion processing in human extrastriate cortex

Background  

Brain imaging and event-related potential studies provide strong evidence that emotional stimuli guide selective attention in visual processing. A reflection of the emotional attention capture is the increased Early Posterior Negativity (EPN) for pleasant and unpleasant compared to neutral images (~150–300 ms poststimulus). The present study explored whether this early emotion discrimination reflects an automatic phenomenon or is subject to interference by competing processing demands. Thus, emotional processing was assessed while participants performed a concurrent feature-based attention task varying in processing demands.     

Recent advances in the simulation of particle-laden flows

A substantial number of algorithms exists for the simulation of moving particles suspended in fluids. However, finding the best method to address a particular physical problem is often highly non-trivial and depends on the properties of the particles and the involved fluid(s) together. In this report, we provide a short overview on a number of existing simulation methods and provide two state of the art examples in more detail. In both cases, the particles are described using a Discrete Element Method (DEM). The DEM solver is usually coupled to a fluid-solver, which can be classified as grid-based or mesh-free (one example for each is given). Fluid solvers feature different resolutions relative to the particle size and separation. First, a multicomponent lattice Boltzmann algorithm (mesh-based and with rather fine resolution) is presented to study the behavior of particle stabilized fluid interfaces and second, a Smoothed Particle Hydrodynamics implementation (mesh-free, meso-scale resolution, similar to the particle size) is introduced to highlight a new player in the field, which is expected to be particularly suited for flows including free surfaces.     

仿选择性注意机制的语音情感识别算法

有效特征的选取一直都是语音情感识别算法的关键。为此,针对语音情感特征选择与构建的问题,一种仿选择性注意机制的语音情感识别算法被提出。考虑到语音信号的时频特性,算法首先计算语音信号的语谱图;其次,模仿选择性注意机制,计算语谱图的颜色、方向和亮度特征图,归一化后形成特征矩阵;然后,将特征矩阵重排列并进行PCA降维,形成情感识别特征向量;最后,利用改进的支持向量机分类方法进行语音情感识别。对愤怒、恐惧、高兴、悲伤和惊奇5种情感的识别实验显示,基于选择性注意的方法能够获得较好的识别效果,平均识别率为85.44%。相比于韵律特征和音质特征,语音情感识别率至少提高10%;相比于其它语谱特征,识别率提高7%左右。      

Quantum Image Encryption Based on Quantum Image Decomposition

Aiming at the slow processing speed of classic image encryption algorithms and the security analysis of existing quantum image encryption algorithms, this paper combines the representation method of quantum images and proposes a quantum image encryption algorithm based on image correlation decomposition. Using the principle of quantum state superposition and measurement, the association between image pixels is established, the image is decomposed into a series of feature sub-images and stored in a complete binary tree set, and different sub-images are operated and encrypted by random phase operation and quantum rotation operation. Then superimpose all the sub-images to obtain the ciphertext image. The algorithm has a larger key space so that it can resist brute force attacks. At the same time, the quantum encryption algorithm has lower computational complexity than classic encryption algorithms. In addition, because the ciphertext image is transmitted in the communication channel in the form of a quantum state, the security of quantum image encryption also surpasses the security of classical image encryption.

     

基于退火遗传算法的少数投影CT图像重建

不完全投影重建算法研究是当今图像重建算法中的一个难点问题，近年来，遗传算法在其中有一定的应用，但还存在一些问题。针对遗传算法在优化中的不足(如早熟收敛及易陷入局部最小等)，将模拟退火机制引入到遗传算子中，提出一种新的算法；并通过选择测试函数，建立将该算法应用于不完全投影图像重建的优化模型，进行计算机模拟。通过对从投影角间距为15°的12个不同投影视角方向获取的投影数据进行分析，结果表明，该算法的重建图像具有较高的精度。