基于降半Г分布函数的自适应遗传算法

针对遗传算法中交叉概率和交异概率难以选取的问题,提出一种新自适应遗传算法;利用降半Г分布函数对交叉概率和变异概率事例上参数随基因串的适合度值而变化。     

参数自适应SMHD滚动轴承IAS信号特征提取方法

针对编码器瞬时角速度(IAS)信号中滚动轴承故障特征提取困难的问题,结合稀疏最大谐波噪声比解卷积(SMHD)算法可在没有先验周期情况下提取信号中周期性脉冲故障分量的优势提出一种参数自适应SMHD滚动轴承IAS信号特征提取方法。首先,利用向前差分法估计IAS信号;然后,利用故障特征(FC)作为自适应选取SMHD优化滤波器长度的评判指标,实现SMHD滤波器长度的自适应确定;再将优化选取的滤波器长度代入SMHD算法对IAS信号进行增强。最后,通过包络分析揭示滚动轴承故障特征。通过对仿真和实测数据进行分析,验证了所提方法的有效性。     

基于多源域领域适应字典学习和稀疏表示的脑电情感识别方法

脑电信号容易记录且不易伪装,基于脑电信号的情感识别越来越受到人们的关注.然而,人类情感具有多样性和个体可变性,基于脑电信号的情感识别仍是情感计算领域的难题.本文提出一种多源域领域适应字典学习和稀疏表示方法.为减少源领域和目标领域数据分布的差异,将所有领域的数据投影到共享子空间,并在共享子空间中学习一个共有字典.根据稀疏重建的最小化类内误差和最大化类间误差准则,稀疏表示具有类别的分辨能力.另外,每个源域自适应学习领域权重,可以避免负迁移的发生.模型参数的求解通过参数交替优化方法,所有参数可同时达到最优解.DEAP数据集的实验结果显示本文方法在所有对比方法中是最优的.     

Structural and Functional Characterization of One Unclassified Glutathione S-Transferase in Xenobiotic Adaptation of Leptinotarsa decemlineata

Arthropod Glutathione S-transferases (GSTs) constitute a large family of multifunctional enzymes that are mainly associated with xenobiotic or stress adaptation. GST-mediated xenobiotic adaptation takes place through direct metabolism or sequestration of xenobiotics, and/or indirectly by providing protection against oxidative stress induced by xenobiotic exposure. To date, the roles of GSTs in xenobiotic adaptation in the Colorado potato beetle (CPB), a notorious agricultural pest of plants within Solanaceae, have not been well studied. Here, we functionally expressed and characterized an unclassified-class GST, LdGSTu1. The three-dimensional structure of the LdGSTu1 was solved with a resolution up to 1.8 Å by X-ray crystallography. The signature motif VSDGPPSL was identified in the “G-site”, and it contains the catalytically active residue Ser14. Recombinant LdGSTu1 was used to determine enzyme activity and kinetic parameters using 1-chloro-2, 4-dinitrobenzene (CDNB), GSH, p-nitrophenyl acetate (PNA) as substrates. The enzyme kinetic parameters and enzyme-substrate interaction studies demonstrated that LdGSTu1 could catalyze the conjugation of GSH to both CDNB and PNA, with a higher turnover number for CDNB than PNA. The LdGSTu1 enzyme inhibition assays demonstrated that the enzymatic conjugation of GSH to CDNB was inhibited by multiple pesticides, suggesting a potential function of LdGSTu1 in xenobiotic adaptation.     

Adaptation techniques for ubiquitous Internet multimedia

We identify the importance of adaptation for the ubiquitous access to Internet multimedia content. The variety of media types combined with the diversity of Internet connection characteristics raises momentous challenges to the achievement of this goal. With adaptation, the characteristics of various media can be adjusted to better match those of the network path and the end device. As a result, adaptive streams enjoy superior robustness and provide substantially better presentation quality than traditional, non‐adaptive ones, especially in variable, capacity‐limited environments. We examine several important factors that influence the design and optimization of the adaptation architecture. First, the location of the adaptation mechanism on the end‐to‐end path must be chosen. Second, the adaptation policy, i.e., the agility in adapting to variations in network path performance, must be specified. Third, the adaptation mechanism needs to be aware of the value of the perceptual quality that the different representations of each medium and the necessary resources to sustain it. Finally, we emphasize the role of supporting mechanisms in enhancing the adaptation process, namely, prioritization, admission control and hand‐off notifications. Based on the aforementioned criteria, we classify existing research approaches and present some of them as case studies. We also describe current commercial solutions and, finally, we discuss future trends in application adaptation in conjunction with recent developments towards wireless access to the Internet. Copyright © 2001 John Wiley & Sons, Ltd.     

Chromatic contrast discrimination: Data and prediction for stimuli varying in L and M cone excitation

Chromatic discrimination data are presented for pulsed and steady stimuli as a function of surround chromaticity and structure. All stimuli and surrounds were at equiluminance and at a constant level of short‐wavelength‐sensitive cone excitation. The test stimulus was a square array of four 1° squares. A 0.07° crosshair of the same chromaticity as the surround separated the squares. Both the test stimuli and the surrounds varied in relative excitation of the long‐wavelength and middle‐wavelength sensitive cones. When stimuli briefly replaced a portion of a steadily viewed background (the Pulse Paradigm), the discriminations were optimal at the background chromaticity and degraded for chromaticities away from the background. The discrimination steps were independent of the background size, which varied from a spatially extensive display to one matching exactly the appearance of the test array. Discrimination was determined only by the spatio‐temporal chromatic contrast of the stimulus relative to the background. When the stimuli were presented continuously within a surround (the Pedestal Paradigm), discrimination was still determined by the surround chromaticity, independent of the surround size. Even a narrow 0.07° crosshair was sufficient to establish optimal discrimination at the crosshair chromaticity. With the surround and crosshair dark, spectral opponent channels maintained an intrinsic normalization near equal energy white. There was little indication of adaptation to the test stimuli. The data were fit by a model of spectral opponency linking detection and discrimination as a function of both retinal illuminance level and chromaticity. The model is explicitly based on observations of the behavior of retinal ganglion cells of the Macaque retina. The model incorporates well‐accepted psychophysical concepts that adaptation in cone spectral‐opponent channels occurs at multiple sites both before and after spectral opponency. © 2000 John Wiley & Sons, Inc. Col Res Appl, 25, 105–115, 2000     

Sub-Carrier and Band Hopped Orthogonal Frequency Division Multiple Access (OFDMA) Systems

This article proposes a novel way of grouping users in an orthogonal frequency division multiplexing (OFDM) communication link based on predefined criteria. The total available spectrum is divided into a number of bands equal to the number of groups. For efficient time-frequency resource allocation, sub-carrier and band hopping are used simultaneously. Under some constraints, sub-carriers can also be assigned to different users based on known channel characteristics using dynamic sub-carrier allocation. Sub-carrier and band hopping are used for mitigating the frequency selectivity of the wireless channel and for minimizing and avoiding interference in the system. The proposed scheme is equally applicable to both downlink and uplink.

PRODUCT GENE REPRESENTATION AND ACQUISITION METHOD BASED ON POPULATION OF PRODUCT CASES

The representation and acquisition of a product gene is a crucial problem in product evolutionary design. A new methodology of product gene representation and acquisition from a population of product cases is proposed, and the methodology for product evolutionary design based on a population of product cases is realized. By properly classifying product cases according to its product species, the populations of product cases are divided and a model is established. Knowledge of the scheme design is extracted and formulated as the function base, principle base, and structure base, which are then combined to form a product gene. Subsequently, the product gene tree is created and represented by object-oriented method. Then combining this method with the evolutionary reasoning technology, an intelligent and automatic evolutionary scheme design of product based on the population of product cases is realized. This design method will be helpful in the processing of knowledge formulation, accumulation, and reuse, and in addressing the difficulty of acquiring design knowledge in traditional design. In addition, the disadvantages of manual case adaptation and update in case-based reasoning can be eliminated. Moreover, by optimizing the design scheme in multiple levels and aspects of product function, principle, and structure etc., the level of creativity in the scheme design can be improved.     

基于FVQ/HMM的无教师说话人自适应

本文提出了一种新的语音识别方法,它综合了VQ、HMM和无教师说话人自适应算法的优点,在每个状态通过用矢量量化误差值取代传统HMM的输出概率值来建立FVQ/HMM,同时采用基于模糊矢量量化的无教师自适应算法,来改变FVQ/HMM的各状态的码字,从而实现对未知说话人的码本适应.本文通过非特定人汉语数码(孤立和连续数码)语音识别实验,把该新的组合方法同基于CHMM的自适应和识别方法进行了比较,实验结果表明该方法的自适应和识别效果优于基于CHMM的方法.     

一种结构感知损失的域适应深度估计方法

针对域适应深度估计方法中域图像间结构差异较大问题,文中提出一种结构感知损失的域适应深度估计方法。该方法通过预训练的卷积神经网络对图像进行特征提取,在特征上进行结构相似性度量,减小了域图像之间的差异性,提高了转换模块的稳定性。该方法使用合成图像深度对和真实图像训练,不需要真实图像的深度标签和物理几何信息。在KITTI数据集上进行实验,深度准确率达到了96.6%,证明该方法可有效提高深度准确率。