基于PCP特征的和弦识别研究与探讨

讨论了音乐识别领域中和弦的四种不同识别方法,给出了基于PCP特征的和弦识别算法.使用PCP作为和弦的特征作为输入送至隐马尔可夫模型中训练,利用Baum-Welch算法估计模型参数,通过Viterbi算法得到正确和弦.通过实验获得了76%的识别率,验证了该算法的可行性.     

粒子群优化模糊神经网络在语音识别中的应用

针对模糊神经网络训练采用BP算法比较依赖于网络的初始条件,训练时间较长,容易陷入局部极值的缺点,利用粒子群优化算法(PSO)的全局搜索性能,将PSO用于模糊神经网络的训练过程.由于基本PSO算法存在一定的早熟收敛问题,引入一种自适应动态改变惯性因子的PSO算法,使算法具有较强的全局搜索能力.将此算法训练的模糊神经网络应用于语音识别中,结果表明,与BP算法相比,粒子群优化的模糊神经网络具有较高的收敛速度和识别率.     

基于改进PROMETHEE的领先用户识别方法

领先用户在技术创新中的作用不容小觑,如何识别出领先用户正逐渐引起人们的关注.提出了一种基于改进PROMETHEE的领先用户识别方法,采用梯形模糊语言形式的属性值,通过结合客观数据与决策者偏好的综合赋权法确定评价指标的权重,并利用基于ULHA算子的群决策方法识别出领先用户.最后通过DHS公司识别领先用户的实例验证了方法的可行性与合理性.     

基于深度学习的电商商品购买意图识别模型

识别用户的购买意图是提升电子商务购买率(PR)的重要方法之一。针对用户购买意图不明确的现象,提出一种新模型。该模型将训练后的Word2Vec(WV)词向量馈入卷积神经网络(CNN),通过深层语义模型(DSSM)进一步提取文本特征。在Keras框架下结合美国建材电商网站家得宝的真实搜索数据进行实证分析。结果表明,在五分类问题中,新模型在测试数据集上的F1-score达80.6%。新模型使用了Word2Vec与CNN提取文本特征,并应用DSSM模型进一步提取了用户检索与商品描述文档在高维空间中的特征表示,最大化利用了用户检索与正确商品描述之间的语义相似度,同时避免了特征提取时主观因素的干扰,提高了商品购买意图的识别效果。     

基于ELM与BP神经网络的新疆雹云预测与识别

冰雹是一种强烈的自然灾害,为了提前对冰雹进行预测,提出了一种BP神经网络做预测,极限学习机做识别的预测方法.方法以雷达回波反射率图像的均值、方差、熵为判别指标,首先训练好能够分类云层特征的极限学习机,其次结合雹云形成过程,使用经过较大样本训练的BP神经网络对未来30分钟数据预测,最后使用经过训练的极限学习机进行云层的类型判别.实验结果表明,BP神经网络能够平均提前5-20分钟预测出降雹,具有较好的预测效果.     

言语知觉反映论

本文对汉语辅音的知觉混淆进行了分析,并对不同语言作了对比研究.在此基础上提出了言语知觉反映论. 研究结果表明:1.对不同的语言,语音特征都是重要的知觉线索,并且物理特性差别较大的语音特征——发音方法,在言语知觉中占优势;2.不同的语言所具有的语音特征系统是不同的;3.语言的社会特性——语言结构对语音的主观识别有着明显的影响.言语的知觉过程是以语音刺激的自然特性为基础,结合语音的社会特性进行的统一过程。 文中还给出了汉语辅音的主要特征及其在不同传递条件下传递信息的能力.通过对比研究看出,汉语音节结构有助于提高音节清晰度.     

基于IFOA-ELM的车载手势识别方法研究

车载手势识别作为一种人机交互方式是提高道路行驶安全性的有效途径.针对传统车载手势识别方法研究中识别准确率和效率较低以及性能不稳定的问题,提出一种改进的果蝇优化算法(IFOA)优化极限学习机(ELM)参数的车载手势识别新方法.首先利用IFOA优化ELM的初始权重w和偏置b;接着采用最佳初始权重和偏置来训练ELM;最后利用IFOA-ELM对提取的车载手势特征向量进行手势类型识别;实验结果表明,与SVM、动态贝叶斯网络、传统ELM、FOA-ELM等分类学习算法相比,方法在高效稳定的前提下取得更高的识别准确率,满足对准确性和实时性要求较高的车载环境中的手势识别.     

基于SVM的车牌数字识别方法

运用支持向量机对车牌字符进行识别,解决了由于图像受客观条件的影响、样本数量不是很大等原因导致的识别率不高的问题.主要针对车牌字符中的数字进行实验,选取了15组数字样本,8组进行训练,7组进行测试,采用交叉验证的思想对SVM进行参数C与g的寻优,并选择合适的核函数,对样本进行训练和预测,对于某些数字的识别率可达到100%,并在相同的训练集和测试集下与BP网络的识别效果进行对比.实验结果表明,SVM在训练样本较少且无字符特征提取的情况下具有很好的识别率,并且有很好的分类推广能力.     

收敛性随机森林模型及其遥感应用

通过引入全局损失函数,提出了一种全局优化的随机森林模型算法,称为θ-β型随机森林,并且利用改进后的模型对城市遥感图进行了检测与识别,识别准确率与识别速率都得到了一定的提高.方法在经典随机森林模型的基础上加入前向反馈模型(Forward Stagewise Additive Model),通过每一层节点的训练结果干预下一层的训练数据(从而改变阈值θ的选择)与训练步长(β),使得最后训练得到的型随机森林收敛速度更快,预测结果更为准确.     

基于遗传算法和模拟退火的插值马氏模型用于真核生物启动子识别

启动子识别是生物信息学领域极具挑战的问题.本文在IMC(Interpolated Markov Chin)的框架下考虑碱基的插入与缺失,采用SA(Simulated Annealing)训练转移概率,以增加模型的鲁棒性,利用GA(Genetic Algorithm)优化IMC插值系数,以克服梯度算法易于陷入局部极值点的缺陷,最后将该模型用于启动子识别,识别率在测试集达到86%.     

基于模糊聚类的语音识别

指出模糊聚类方法影响语间识别的正确率 ,提出用基于分解构造的模糊聚类方法设立语音模式的参考向量 ,说明该方法能提高语章识别的正确率。     

影响电子商务客户流失因素分析

鉴于客户流失给电子商务带来的严重损失,需要深入分析影响客户流失的因素,从而指导企业有针对性地改进服务.通过语言信息理论,邀请专家对电子商务客户流失因素进行了分析.通过基于正态分布的有序加权平均(Ordered Weighted Averaging,OWA)对结果进行了处理.在此基础上,提出了影响客户流失因素的分析方法,为企业改进服务提供了强有力的智力支持.     

A multilevel image thresholding using the honey bee mating optimization

Image thresholding is an important technique for image processing and pattern recognition. Many thresholding techniques have been proposed in the literature. Among them, the maximum entropy thresholding (MET) has been widely applied. In this paper, a new multilevel MET algorithm based on the technology of the honey bee mating optimization (HBMO) is proposed. This proposed method is called the maximum entropy based honey bee mating optimization thresholding (MEHBMOT) method. Three different methods such as the particle swarm optimization (PSO), the hybrid cooperative-comprehensive learning based PSO algorithm (HCOCLPSO) and the Fast Otsu’s method are also implemented for comparison with the results of the proposed method. The experimental results manifest that the proposed MEHBMOT algorithm can search for multiple thresholds which are very close to the optimal ones examined by the exhaustive search method. In comparison with the other three thresholding methods, the segmentation results using the MEHBMOT algorithm is the best and its computation time is relatively low. Furthermore, the convergence of the MEHBMOT algorithm can rapidly achieve and the results validate that the proposed MEHBMOT algorithm is efficient.     

Interpretation of fuzzy data by means of fuzzy rules with applications to speech recognition

Recently, a speech recognition methodology has been proposed which has as one of its main principles the explicit assumption of intrinsic uncertainty of the data (speech signals) and inexactness of the knowledge (acoustic phonetic, etc…) available to interpret them. The main problem presented by this methodology is that of parsing ‘fuzzy data’ by means of ‘fuzzy rules’. To solve this problem, an appropriate fuzzy parsing and interpretation scheme has been proposed. It assumes the data to be represented as strings of ‘fuzzy symbols’, defined as fuzzy sets over the appropriate set of categories, and knowledge as finite-state networks with the arcs labelled by fuzzy symbols of the same type. A formal presentation of this scheme is the main topic of this paper. Included is a brief discussion of the application to Automatic Speech Recognition, and a summary of some results obtained from an implementation example.     

Pattern classification by using improved wavelet Compressed Zernike Moments

In this paper, an improved Feature Extraction Method (FEM), which selects discriminative feature sets able to lead to high classification rates in pattern recognition tasks, is presented. The resulted features are the wavelet coefficients of an improved compressed signal, consisting of the Zernike moments amplitudes. By applying a straightforward methodology, it is aimed to construct optimal feature vectors in the sense of vector dimensionality and information content for classification purposes. The resulting surrogate feature vector is of lower dimensionality than the original Zernike moment feature vector and thus more appropriate for pattern recognition tasks.Appropriate validation tests have been arranged, in order to investigate the performance of the proposed algorithm by measuring the discriminative power of the new feature vectors despite the information loss.     

农作物单产的最大熵分布及在费率厘定上的应用

农作物单产分布的确定是农业保险中费率厘定的基础。本文引入最大熵原理,基于最大熵优化模型得出农作物单产的最大熵分布,并以此进行费率厘定。同时以辽宁省主要作物水稻、玉米、大豆和花生为例,确定了该四种农作物的费率,分别为4.45%、6.77%、6.34%、6.43%。结果表明:利用最大熵分布理论进行费率厘定不需要事先假定农作物单产分布的形式,而且考虑了更多作物单产分布的信息,为农业保险费率的合理精算提供一种新的可供选择的方法,有助于农业风险决策的科学化。     

An adaptive cellular updating scheme for the continuous–discontinuous cellular automaton method

The updating scheme is very important for cellular automata, which costs the vast majority of CPU time of the whole calculating process and dramatically influences the convergence rate, and the traditional node sequence scheme is often ineffective. Aimed at this issue, based on the unbalanced nodal force, an adaptive updating scheme for continuous–discontinuous cellular automaton method is proposed in this paper. The cell whose unbalanced force is maximum is the first updating cell, and the next one is its neighbors whose unbalanced force is also maximum in all remain neighbors, and do the updating operation according to this scheme until the updating for all cells is finished. Theory and convergence of the present scheme is studied, then we can get the maximum calculating efficiency, and the computational expanse can be greatly improved. Some examples by different iteration schemes are shown to demonstrate that the present scheme is efficient and time saving.     

基于粗糙数和信息熵的产品设计方案灰色关联评估模型

针对复杂产品方案设计中指标属性信息的不完全性和不确定性,研究了一种基于粗糙数和信息熵理论的灰色关联评估模型。首先通过引入粗糙数序列的范数实现粗糙数评估矩阵的规范化处理,并利用熵权对指标属性值进行权重集结,然后构建理想最优特征序列,并借助基于信息还原算子的粗糙相似关联度来获得最优评估方案。信息熵赋权可减少主观赋权产生的人为因素影响,信息还原算子可避免评估过程中的信息失真现象。最后通过工程机械产品的实例,验证了该评估模型的有效性和实用性,便于对复杂产品的方案设计进行评估和优选。     

“货到人”拣选系统中电商订单分批优化研究

本文在电商行业迅速发展及人工智能日趋成熟的背景下,研究以自动导引车(AGV)为搬运工具的“货到人”拣选系统订单分批问题。首先以最小化AGV搬运货架次数为目标建立订单分批模型,然后提出了基于货架相似度的两阶段订单分批算法,在第一阶段分为新批次创建及订单加入批次两个步骤得到初始解;在第二阶段采用局部搜索改进初始解。在算法中构造订单选取货架规则,定义货架相似度函数并设计两种方法创建新批次,同时考虑周转箱数量限制求解订单分批方案。最后通过实验测试验证了模型和算法的有效性,分析了两种批次创建方法的适用性,并通过灵敏度分析给出了合理的周转箱数量配置建议。本研究可为采用“货到人”拣选系统的企业通过订单分批优化进一步提高AGV拣选效率提供理论依据和实践指导。     

Particle hydrodynamics based on maximum entropy approximation

In this contribution, we propose a new particle based scheme for computational hydrodynamics. In contrast to existing particle based approaches, we use a maximum entropy approximation scheme to construct corresponding shape functions. These allow first order consistency as well as an efficient realisation of essential boundary conditions. The scheme is relative simple from an algorithmic point of view and computations can be efficiently accelerated using graphic processing units. (© 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim)