Emotion recognition based on convolutional neural networks and heterogeneous bio-signal data sources

Emotion recognition is a crucial application in human–computer interaction. It is usually conducted using facial expressions as the main modality, which might not be reliable. In this study, we proposed a multimodal approach that uses 2-channel electroencephalography (EEG) signals and eye modality in addition to the face modality to enhance the recognition performance. We also studied the use of facial images versus facial depth as the face modality and adapted the common arousal–valence model of emotions and the convolutional neural network, which can model the spatiotemporal information from the modality data for emotion recognition. Extensive experiments were conducted on the modality and emotion data, the results of which showed that our system has high accuracies of 67.8% and 77.0% in valence recognition and arousal recognition, respectively. The proposed method outperformed most state-of-the-art systems that use similar but fewer modalities. Moreover, the use of facial depth has outperformed the use of facial images. The proposed method of emotion recognition has significant potential for integration into various educational applications.     

基于集成BP网络的人脸识别研究

在对人脸图像使用小波变换进行数据压缩的基础上,使用PCA进行特征提取,再将特征输入集成BP神经网络实现对人脸图像的识别。集成BP网络将多分类问题转换为多个相互独立的二分类问题,在提高网络泛化能力的同时缩短了网络的训练时间。另外,集成网络通过增添子网络或者重新训练子网络的方法解决了网络"失忆"问题,使其具有增量式学习的能力。通过在ORL人脸库上仿真的实验,证明了集成网络的人脸识别以及增量学习都具有良好的性能。     

Web proxy cache replacement scheme based on back-propagation neural network

Web proxy caches are used to reduce the strain of contemporary web traffic on web servers and network bandwidth providers. In this research, a novel approach to web proxy cache replacement which utilizes neural networks for replacement decisions is developed and analyzed. Neural networks are trained to classify cacheable objects from real world data sets using information known to be important in web proxy caching, such as frequency and recency. Correct classification ratios between 0.85 and 0.88 are obtained both for data used for training and data not used for training. Our approach is compared with Least Recently Used (LRU), Least Frequently Used (LFU) and the optimal case which always rates an object with the number of future requests. Performance is evaluated in simulation for various neural network structures and cache conditions. The final neural networks achieve hit rates that are 86.60% of the optimal in the worst case and 100% of the optimal in the best case. Byte-hit rates are 93.36% of the optimal in the worst case and 99.92% of the optimal in the best case. We examine the input-to-output mappings of individual neural networks and analyze the resulting caching strategy with respect to specific cache conditions.     

基于角微特征的神经网络识别人脸特征

提出一种利用人脸角微特征几何特性的图像预处理,建立BP神经网络识别人脸特征模型的方法。研究了角微特征提取和具体算法,讨论了BP网络结构的设计,输入、输出层设计和隐藏层节点选取问题。微特征提取,可以降低网络输入维度,对于识别不同角度、不同表情的人脸图像提供了可能性。利用ORL人脸图像数据库做实验,结果表明此方法有效。     

Automatic inspection system of LED chip using two-stages back-propagation neural network

This study proposed an automatic LED defect detection system to investigate the defects of LED chips. Such defects include fragment chips, scratch marks and remained gold on the pad area, scratch marks on the luminous zone, and missing luminous zone respectively. The system was based on positioning and image acquisition, appearance feature recognition, and defect classification. The normalized correlation coefficient method was used to locate the chip and acquire its image, the K-means clustering method was used to distinguish the appearance, pad area, and luminous zone of chips. In terms of pad area detection, histogram equalization was used to enhance the pad image contrast, and statistical threshold selection and morphological closing were applied to modify the impure points in the pad. Feature values of the pad area were then calculated. The optimal statistical threshold separated the luminous zone and background from the substrate. After processed with closing operation, features of the luminous zone were extracted. Finally, features of each part were clarified by an efficient two-step back-propagation neural network, where a designed appearance classifier and an internal structure classifier were used for recognition. From experiments, total recognition rate of this study achieved 97.83 %, proving that the detection method proposed by this study can efficiently detect LED chip defects.     

基于PSO和BP复合算法的模糊神经网络控制器

为了克服单独应用粒子群算法（PSO）或BP算法训练模糊神经网络控制器参数时存在的缺陷,提出了一种训练模糊神经网络参数的PSO＋BP算法。该算法将二者相结合,即在PSO算法中加入一个BP算子,以充分利用PSO算法的全局寻优能力和BP算法的局部搜索能力,从而更有效地提高其收敛速度、训练效率和提高该模糊神经网络控制器的控制效果。最后的仿真实验结果验证了该基于PSO＋BP复合算法的模糊神经网络控制器的有效性和可行性。     

A three-layer back-propagation neural network for spam detection using artificial immune concentration

In this paper, a three-layer back-propagation neural network (BPNN) is employed for spam detection by using a concentration based feature construction (CFC) approach. In the CFC approach, ‘self’ and ‘non-self’ concentrations are constructed through ‘self’ and ‘non-self’ gene libraries, respectively, to form a two-element concentration vector for expressing the e-mail efficiently. A three-layer BPNN with two-element input is then employed to classify e-mails automatically. Comprehensive experiments are conducted on two public benchmark corpora PU1 and Ling to demonstrate that the proposed CFC approach based BPNN classifier not only has a very much fast speed but also achieves 97 and 99% of classification accuracy on corpora PU1 and Ling by just using a two-element concentration feature vector.     

基于循环卷积神经网络的实体关系抽取方法研究

针对目前大多数关系抽取中对于文本语料中较长的实体共现句,往往只能获取到局部的特征,并不能学习到长距离依赖信息的问题,提出了一种基于循环卷积神经网络与注意力机制的实体关系抽取模型。将擅长处理远距离依赖关系的循环神经网络GRU加入到卷积神经网络的向量表示阶段,通过双向GRU学习得到词语的上下文信息向量,在卷积神经网络的池化层采取分段最大池化方法,在获取实体对结构信息的同时,提取更细粒度的特征信息,同时在模型中加入基于句子级别的注意力机制。在NYT数据集的实验结果表明提出方法能有效提高实体关系抽取的准确率与召回率。     

基于多通道卷积神经网的实体关系抽取

针对实体关系抽取任务中,传统基于统计学习的方法构建特征费时费力、现有深度学习方法依赖单一词向量的表征能力的问题,提出多通道卷积神经网模型。首先使用不同的词向量将输入语句进行映射,作为模型不同通道的输入;然后使用卷积神经网自动提取特征;最后通过softmax分类器输出关系类型,完成关系抽取任务。和其他模型相比,该模型可以获取输入语句更丰富的语义信息,自动学习出更具有区分度的特征。在SemEval-2010 Task 8 数据集上的实验结果表明提出的多通道卷积神经网模型较使用单一词向量的模型更适合处理关系抽取任务。     

A multi-industry bankruptcy prediction model using back-propagation neural network and multivariate discriminant analysis

The accurate prediction of corporate bankruptcy for the firms in different industries is of a great concern to investors and creditors, as the reduction of creditors’ risk and a considerable amount of saving for an industry economy can be possible. This paper presents a multi-industry investigation of the bankruptcy of Korean companies using back-propagation neural network (BNN). The industries include construction, retail, and manufacturing. The study intends to suggest the industry specific model to predict bankruptcy by selecting appropriate independent variables. The prediction accuracy of BNN is compared to that of multivariate discriminant analysis.The results indicate that prediction using industry sample outperforms the prediction using the entire sample which is not classified according to industry by 6–12%. The prediction accuracy of bankruptcy using BNN is greater than that of MDA. The study suggests insights for the practical industry model for bankruptcy prediction.     

基于BP神经网络的专利自动分类方法

提出了一种基于后向传播神经网络的专利自动分类方法.通过中文分词从专利文件集中提取特征项,并根据特征项在专利文件中出现的频率赋予其权重,从而将每篇专利文件表示为一个特征项向量.为取得较好的BP神经网络(BPN)训练效果,使用X2统计方法进行特征向量降维,并使用BPN专利分类器进行专利文件分类.用国际分类号为H02下的专利文件作为测试数据,取得了较好的分类效果.     

基于两级神经网络的连续哼唱特征提取

针对用户以任意字词连续哼唱的情况下,哼唱特征提取中音符分割、音符识别难度大的问题,提出了一种基于两级神经网络的哼唱特征提取方法。第一级采用BP神经网络实现哼唱音符分割,得到独立音符;第二级采用RBF神经网络识别分割出的各个音符,获得音符的MIDI音高值。实验结果表明,该方法能较好地完成哼唱特征的提取,适合于实际哼唱检索系统对连续哼唱的要求。     

基于图神经网络的联合实体关系抽取

从非结构化文本中联合提取实体和关系是信息抽取中的一项重要任务。现有方法取得了可观的性能,但仍受到一些固有的限制,如错误传播、预测存在冗余性、无法解决关系重叠问题等。为此,提出一种基于图神经网络的联合实体关系抽取模型BSGB（BiLSTM+SDA-GAT+BiGCN）。BSGB分为两个阶段：第一阶段将语义依存分析扩展到语义依存图,提出融合语义依存图的图注意力网络（SDA-GAT）,通过堆叠BiLSTM和SDA-GAT提取句子序列和局部依赖特征,并进行实体跨度检测和初步的关系预测;第二阶段构建关系加权GCN,进一步建模实体和关系的交互,完成最终的实体关系三元组抽取。在NYT数据集上的实验结果表明,该模型F₁值达到了67.1%,对比在该数据集的基线模型提高了5.2%,对重叠关系的预测也有大幅改善。     

基于多二维RBF神经网络的航空雷达信号分选

针对航空雷达信号分选中侦察装备普遍存在的信号分选实时性差,分选结果经常出现增批、漏批现象的缺点.为了提高侦察系统在复杂电磁环境下准确快速的分选出雷达辐射源信号,根据径向基(RBF)神经网络通过理想数据训练后能够对未知数据进行分类的特点,将径向基神经网络算法用于对航空雷达侦察信号的分选,在此基础上提出了一种新型多二维径向基神经网络结构,通过与BP网络、RBF网络的对比,多二维径向基神经网络的识剐率优于其它几种网络,而且其结构便于实现.通过试验结果可以得出,多二维径向基神经网络能够提高雷迭信号分选的准确率.     

多视角神经网络非接触式脉搏信号提取

目的 远程光体积描记（remote photoplethysmography,rPPG）是一种基于视频的非接触式心率测量技术,受到学者的广泛关注。从视频数据中提取脉搏信号需要同时考虑时间和空间信息,然而现有方法往往将空间处理与时间处理割裂开,从而造成建模不准确、测量精度不高等问题。本文提出一种基于多视角2维卷积的神经网络模型,对帧内和帧间相关性进行建模,从而提高测量精度。方法 所提网络包括普通2维卷积块和多视角卷积块。普通2维卷积块将输入数据在空间维度做初步抽象。多视角卷积块包括3个通道,分别从输入数据的高—宽、高—时间、宽—时间3个视角进行2维卷积操作,再将3个视角的互补时空特征进行融合得到最终的脉搏信号。所提多视角2维卷积是对传统单视角2维卷积网络在时间维度的扩展。该方法不破坏视频原有结构,通过3个视角的卷积操作挖掘时空互补特征,从而提高脉搏测量精度。结果 在公共数据集PURE（pulse rate detection dataset）和自建数据集Self-rPPG（self-built rPPG dataset）上的实验结果表明,所提网络提取脉搏信号的信噪比相比于传统方法在两个数据集上分别提高了3.92 dB和1.92 dB,平均绝对误差分别降低了3.81 bpm和2.91 bpm;信噪比相比于单视角网络分别提高了2.93 dB和3.20 dB,平均绝对误差分别降低了2.20 bpm和3.61 bpm。结论 所提网络能够在复杂环境中以较高精度估计出受试者的脉搏信号,表明了多视角2维卷积在rPPG脉搏提取的有效性。与基于单视角2维神经网络的rPPG算法相比,本文方法提取的脉搏信号噪声、低频分量更少,泛化能力更强。     

基于三维卷积神经网络的湖泊提取算法

针对现有分析湖泊几何信息算法的二维图像湖泊轮廓提取精度低的问题,提出了一种基于三维卷积神经网络的湖泊提取算法。首先,基于平整度信息从激光扫描点云中定位出候选湖泊并对输入的候选区域点云进行体素化组织,作为神经网络的输入;同时,通过深度学习技术,从候选区域中过滤非湖泊区域;然后,基于方向链码算法从点云中提取湖泊的边缘并分析其几何形状信息。实验结果表明,所提算法在提取激光扫描点云中的湖泊精度可达到96.34%,与当前在二维图像中的湖泊提取算法相比,可对目标湖泊形状信息进行计算与分析,从而为湖泊监测与管理提供方便。     

Integration of fuzzy neural network and artificial immune system-based back-propagation neural network for sales forecasting using qualitative and quantitative data

Sales forecasting plays a very important role in business operation. Many researches generally employ statistical methods, such as regression or auto-regressive integrated moving average model, to forecast the product sales. However, they only can consider the quantitative data. Some exogenous qualitative variables have more influence on forecasting result. Thus, this study attempts to propose a integrated forecasting system which is able to consider both quantitative and qualitative factors to achieve a more comprehensive result. Basically, fuzzy neural network is first employed to capture the expert knowledge regarding some qualitative factors. Then, it is combined with the time series data using an artificial immune system based back-propagation neural network. A laptop sales data set provided by a distributor in Taiwan is applied to verify the proposed approach. The computational result indicates that the proposed approach is superior to other forecasting methods. It can be used to decrease the inventory costs and enhance the customer satisfaction.     

A modified Smith predictive scheme based back-propagation neural network approach for FOPDT processes control

In this paper, a novel control scheme to deal with process uncertainties in the form of disturbance loads and modelling errors, as well as time-varying process parameters is proposed by applying the back-propagation neural network (BPNN) approach. A BPNN predictive controller that replaces the entire Smith predictor structure is initially trained offline. Lyapunov direct method is used to prove that the convergence of this BPNN is guaranteed by selecting a suitable learning rate during the learning process. However, the Smith predictor based BPNN control is an off-line training based algorithm, which is a time consuming method and requires a known process plant input from the controller. A desired control input to the process is difficult to obtain for the training of the network. As a result a group of proper training data (target control inputs and outputs) can hardly be provided. In order to overcome this problem, a BPNN with an on-line training algorithm is introduced for the control of a First Order plus Dead Time (FOPDT) process. The stability analysis is carried out using the Lyapunov criterion to demonstrate the network convergence ability. Simulation results show that this proposed online trained neural Smith predictor based controller provides excellent robustness to process modelling errors and disturbance loads, and high adaptability to time varying processes parameters.     

A user authentication system using back-propagation network

Information security has been a critical issue in the field of information systems. One of the key factors in the security of a computer system is how to identify the authorization of users. Password-based user authentication is widely used to authenticate a legitimate user in the current system. In conventional password-based user authentication schemes, a system has to maintain a password table or verification table which stores the information of users IDs and passwords. Although the one-way hash functions and encryption algorithms are applied to prevent the passwords from being disclosed, the password table or verification table is still vulnerable. In order to solve this problem, in this paper, we apply the technique of back-propagation network instead of the functions of the password table and verification table. Our proposed scheme is useful in solving the security problems that occurred in systems using the password table and verification table. Furthermore, our scheme also allows each user to select a username and password of his/her choice.