Artificial neural networks for pattern recognition

This tutorial article deals with the basics of artificial neural networks (ANN) and their applications in pattern recognition. ANN can be viewed as computing models inspired by the structure and function of the biological neural network. These models are expected to deal with problem solving in a manner different from conventional computing. A distinction is made between pattern and data to emphasize the need for developing pattern processing systems to address pattern recognition tasks. After introducing the basic principles of ANN, some fundamental networks are examined in detail for their ability to solve simple pattern recognition tasks. These fundamental networks together with the principles of ANN will lead to the development of architectures for complex pattern recognition tasks. A few popular architectures are described to illustrate the need to develop an architecture specific to a given pattern recognition problem. Finally several issues that still need to be addressed to solve practical problems using ANN approach are discussed. This paper is mostly a consolidation of work reported by several researchers in the literature, some of which is cited in the references. The author has borrowed several ideas and illustrations from the references quoted in this paper.     

Multi-step ART1 algorithm for recognition of defect patterns on semiconductor wafers

The integrated circuits (ICs) on wafers are highly vulnerable to defects generated during the semiconductor manufacturing process. The spatial patterns of locally clustered defects are likely to contain information related to the defect generating mechanism. For the purpose of yield management, we propose a multi-step adaptive resonance theory (ART1) algorithm in order to accurately recognise the defect patterns scattered over a wafer. The proposed algorithm consists of a new similarity measure, based on the p-norm ratio and run-length encoding technique and pre-processing procedure: the variable resolution array and zooming strategy. The performance of the algorithm is evaluated based on the statistical models for four types of simulated defect patterns, each of which typically occurs during fabrication of ICs: random patterns by a spatial homogeneous Poisson process, ellipsoid patterns by a multivariate normal, curvilinear patterns by a principal curve, and ring patterns by a spherical shell. Computational testing results show that the proposed algorithm provides high accuracy and robustness in detecting IC defects, regardless of the types of defect patterns residing on the wafer.     

基于卷积神经网络和Transformer网络的鸟声识别

针对传统鸟声识别算法中特征提取方式单一、分类识别准确率低等问题,提出一种结合卷积神经网络和Transformer网络的鸟声识别方法。该方法综合考虑网络局部特征学习和全局上下文依赖性构造,从原始鸟声音频信号中提取短时傅里叶变换(Short Time Fourier Transform,STFT)语谱图特征,将其输入到卷积神经网络(ConvolutionalNeural Network,CNN)中提取局部频谱特征信息,同时提取鸟声信号的对数梅尔特征及一阶差分、二阶差分特征用于合成梅尔频率倒谱系数(Mel Frequency Cepstrum Coefficient,MFCC)混合特征向量,将其输入到Transformer网络中获取全局序列特征信息,最后融合所提取的特征可得到更丰富的鸟声特征参数,通过Softmax分类器得到鸟声识别结果。在Birdsdata和xeno-canto鸟声数据集上进行实验,平均识别准确率分别达到了97.81%和89.47%。实验结果表明该方法相较于其他现有的鸟声识别模型具有更高的识别准确率。     

基于改进一维卷积神经网络的滚动轴承故障识别

滚动轴承的故障识别对于防止旋转机械系统故障恶化并保证其安全运行具有重要意义。针对现有智能诊断模型参数多、识别效率低的问题,提出一种基于改进一维卷积神经网络的滚动轴承故障识别(FRICNN–1D)方法。通过引入1×1卷积核增强一维卷积神经网络模型的非线性表达能力;并用全局平局池化层代替传统卷积神经(CNN)网络中的全连接层,以降低模型参数和计算量,且防止过拟合现象。试验结果表明,该方法可以准确识别滚动轴承不同故障状态,具有一定的工程实际应用潜力。     

基于粗糙集和神经网络的联合模式识别系统

提出了一种划分属性离散区间的新方法．针对这种划分，提出一种约简和去噪的方法．随后，建立了粗糙集和LVQ神经网络的联合模式识别系统．最后，比较了用该系统和仅用神经网络进行识别的效果，证明了该方法的有效性．     

Content‐based gastric image retrieval using convolutional neural networks

The endoscopy procedure has demonstrated great efficiency in detecting stomach lesions, with extensive numbers of endoscope images produced globally each day. The content‐based gastric image retrieval (CBGIR) system has demonstrated substantial potential in gastric image analysis. Gastric precancerous diseases (GPD) have higher prevalence in gastric cancer patients. Thus, effective intervention is crucial at the GPD stage. In this paper, a CBGIR method is proposed using a modified ResNet‐18 to generate binary hash codes for a rapid and accurate image retrieval process. We tested several popular models (AlexNet, VGGNet and ResNet), with ResNet‐18 determined as the optimum option. Our proposed method was valued using a GPD data set, resulting in a classification accuracy of 96.21 ± 0.66% and a mean average precision of 0.927 ± 0.006 , outperforming other state‐of‐art conventional methods. Furthermore, we constructed a Gastric‐Map (GM) based on feature representations in order to visualize the retrieval results. This work has great auxiliary significance for endoscopists in terms of understanding the typical GPD characteristics and improving aided diagnosis.     

Pre-trained convolutional neural networks as feature extractors for diagnosis of breast cancer using histopathology

Several researchers are trying to develop different computer-aided diagnosis system for breast cancer employing machine learning (ML) methods. The inputs to these ML algorithms are labeled histopathological images which have complex visual patterns. So, it is difficult to identify quality features for cancer diagnosis. The pre-trained Convolutional Neural Networks (CNNs) have recently emerged as an unsupervised feature extractor. However, a limited investigation has been done for breast cancer recognition using histopathology images with CNN as a feature extractor. This work investigates ten different pre-trained CNNs for extracting the features from breast cancer histopathology images. The breast cancer histopathological images are obtained from publicly available BreakHis dataset. The classification models for the different feature sets, which are obtained using different pre-trained CNNs in consideration, are developed using a linear support vector machine. The proposed method outperforms the other state of art methods for cancer detection, which can be observed from the results obtained.     

Landmark-based multi-region ensemble convolutional neural networks for bone age assessment

Bone age assessment based on hand X-ray imaging is important in pediatry medicine. At present, prediction of bone age is mainly performed by the manual comparison with the existing atlas. To develop an automatic regression framework based on deep learning with high performance and efficiency. A landmark-based multi-region convolutional neural networks for automatic bone age assessment based on left hand X-ray images was proposed. The deep alignment network localized multiple landmarks distributed over the hand, and cropped the local regions to establish the multi-region ensemble convolutional neural networks with different sub-network combinations. The modified loss function and the optimized bone sub-regions were applied to train the networks. The experiments on Digital Hand Atlas Database revealed that the mean absolute error of bone age assessment was 0.52 ± 0.25 years. It is the first study to predict bone age using deep learning methods throughout the entire process including image preprocessing, landmark localization and bone age predication. The proposed method outperformed most of the existing state-of-the-art deep learning methods and achieved good results compared with the expert's experience. It can improve the efficiency of the medical doctors while minimizing the subjective errors.     

基于CNN的点云图像融合目标检测

针对自动驾驶场景中目标检测存在尺度变化、光照变化和缺少距离信息等问题,提出一种极具鲁棒性的多模态数据融合目标检测方法,其主要思想是利用激光雷达提供的深度信息作为附加的特征来训练卷积神经网络(CNN)。首先利用滑动窗对输入数据进行切分匹配网络输入,然后采用两个CNN特征提取器提取RGB图像和点云深度图的特征,将其级联得到融合后的特征图,送入目标检测网络进行候选框的位置回归与分类,最后进行非极大值抑制(NMS)处理输出检测结果,包含目标的位置、类别、置信度和距离信息。在KITTI数据集上的实验结果表明,本文方法通过多模态数据的优势互补提高了在不同光照场景下的检测鲁棒性,附加滑动窗处理改善了小目标的检测效果。对比其他多种检测方法,本文方法具有检测精度与检测速度上的综合优势。     

Smart connected electronic gastroscope system for gastric cancer screening using multi-column convolutional neural networks

Gastroscopy is a widely adopted method for gastric cancer screening and early diagnosis. Clinical studies show that it can effectively prolong patient life and maximise therapeutic effect. However, it is difficult for doctors to identify and detect lesions in real time, which manifests as the major challenge in gastroscopy. In this paper, we propose SCEG, a smart connected electronic gastroscopy system that performs dynamic cancer screening in gastroscopy. By integrating electronic gastroscopy with cloud-based medical image analysis service, we develop an AdaBoost-based multi-column convolutional neural network (MCNN) for enhancing gastric cancer screening. Experimental results show that the proposed MCNN approach significantly outperforms other competing approaches.     

Deep convolutional neural networks for eigenvalue problems in mechanics

We show that deep convolutional neural networks (CNNs) can massively outperform traditional densely connected neural networks (NNs) (both deep or shallow) in predicting eigenvalue problems in mechanics. In this sense, we strike out in a new direction in mechanics computations with strongly predictive NNs whose success depends not only on architectures being deep but also being fundamentally different from the widely used to date. We consider a model problem: predicting the eigenvalues of one-dimensional (1D) and two-dimensional (2D) phononic crystals. For the 1D case, the optimal CNN architecture reaches 98% accuracy level on unseen data when trained with just 20 000 samples, compared to 85% accuracy even with 100 000 samples for the typical network of choice in mechanics research. We show that, with relatively high data efficiency, CNNs have the capability to generalize well and automatically learn deep symmetry operations, easily extending to higher dimensions and our 2D case. Most importantly, we show how CNNs can naturally represent mechanical material tensors, with its convolution kernels serving as local receptive fields, which is a natural representation of mechanical response. Strategies proposed are applicable to other mechanics' problems and may, in the future, be used to sidestep cumbersome algorithms with purely data-driven approaches based upon modern deep architectures.     

Multiview convolutional neural networks for lung nodule classification

To find a better way to screen early lung cancer, motivated by the great success of deep learning, we empirically investigate the challenge of classifying lung nodules in computed tomography (CT) in an end‐to‐end manner. Multi‐view convolutional neural networks (MV‐CNN) are proposed in this article for lung nodule classification. Unlike the traditional CNNs, a MV‐CNN takes multiple views of each entered nodule. We carry out a binary classification (benign and malignant) and a ternary classification (benign, primary malignant, and metastatic malignant) using the Lung Image Database Consortium and Image Database Resource Initiative database. The results show that, for binary or ternary classifications, the multiview strategy produces higher accuracy than the single view method, even for cases that are over‐fitted. Our model achieves an error rate of 5.41 and 13.91% for binary and ternary classifications, respectively. Finally, the receiver operating characteristic curve and t‐distributed stochastic neighbor embedding algorithm are used to analyze the models. The results reveal that the deep features learned by the model proposed in this article have a higher separability than features from the image space and the multiview strategies; therefore, researchers can get better representation. © 2017 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 27, 12–22, 2017     

基于噪声和图卷积神经网络的电机故障诊断

文章针对旋转机械设备维护和噪声监测治理的需求,提出了一种基于电机噪声信号和图卷积神经网络的故障诊断算法。该算法对时域数据进行傅里叶变换,将变换后的频域数据转化为图数据,利用提出的新型图卷积神经网络结构对图数据进行训练并分类。搭建电机故障实验平台,完成了6种不同状态的电机噪声信号采集与实验验证。实验结果表明,图卷积神经网络能根据有限的电机噪声信号有效识别出电机故障,并具有一定的小样本学习能力,能够在样本量较少的情况下进行故障分类。对比分析表明,该算法分类准确率优于K最近邻-图算法、一维卷积神经网络、自动编码器和支持向量机等其他算法,为实际工程应用提供了参考。     

结合多尺度卷积网络和双端注意力机制的水声目标识别

水声目标识别是被动声呐系统的主要应用之一。为了进一步提升小样本条件下水下目标的识别率,文章提出一种基于多尺度卷积和双端注意力机制相融合的方法。首先,提取梅尔倒谱系数,色度谱和计算谱对比度等特征,建立基于多类别特征子集的三维聚合特征。其次,采用多尺度卷积滤波器算子构造多分辨率卷积神经网络,以更好地适应三维聚合特征的时频结构。另外,采用双端注意力模型捕获样本的全局依赖和局部特性。采用基于指数加权的对数交叉熵函数作为损失函数,提升样本数较少类别的识别率。实验结果表明,该方法在ShipsEar数据上的平均识别率为95.5%,取得了较好的分类效果。     

基于U型全卷积神经网络的路面裂缝检测

路面裂缝检测是道路运营和维护的一项重要工作,由于裂缝没有固定形状而且纹理特征受光照影响大,基于图像的精确裂缝检测是一项巨大的挑战。本文针对裂缝图像的特点,提出了一种U型结构的卷积神经网络UCrackNet。首先在跳跃连接中加入Dropout层来提高网络的泛化能力;其次,针对上采样中容易产生边缘轮廓失真的问题,采用池化索引对图像边界特征进行高保真恢复;最后,为了更好地提取局部细节和全局上下文信息,采用不同扩张系数的空洞卷积密集连接来实现感受野的均衡,同时嵌入多层输出融合来进一步提升模型的检测精度。在公开的道路裂缝数据集CrackTree206和AIMCrack上测试表明,该算法能有效地检测出路面裂缝,并且具有一定的鲁棒性。     

Lymph node detection method based on multisource transfer learning and convolutional neural network

Recently years, convolutional neural networks (CNNs) have proven to be powerful tools for a broad range of computer vision tasks. However, training a CNN from scratch is difficult because it requires a large amount of labeled training data, which remains a challenge in medical imaging domain. To this end, deep transfer learning (TL) technique is widely used for many medical image tasks. In this paper, we propose a novel multisource transfer learning CNN model for lymph node detection. The mechanism behind it is straightforward. Point-wise (1 × 1) convolution is used to fuse multisource transfer learning knowledge. Concretely, we view the transferred features as priori domain knowledge and 1 × 1 convolutional operation is implemented after pre-trained convolution layers to adaptively combine the transfer information for target task. In order to learn non-linear transferred features and prevent over-fitting, we present an encode process for the pre-trained convolution kernels. At last, based on convolutional factorization technique, we train the proposed CNN model and the encoder process jointly, which improves the feasibility of our approach. The effectiveness of the proposed method is verified on lymph node (LN) dataset: 388 mediastinal LNs labeled by radiologists in 90 patient CT scans, and 595 abdominal LNs in 86 patient CT scans for LN detection. Our method demonstrates sensitivities of about 85%/71% at 3 FP/vol. and 92%/85% at 6 FP/vol. for mediastinum and abdomen respectively, which compares favorably to previous methods.     

基于小波包时频图特征和卷积神经网络的水声信号分类

水下声信号分类是水声学研究的一个重要方向.一个有效的特征提取和分类决策方法对水声信号分类技术至关重要.文章将鱼声、商船辐射噪声和风关噪声三类实测的水声信号在小波包分解的基础上提取时频图特征,并搭建了一个七层结构的卷积神经网络作为分类器.研究结果表明:三种水声信号的小波包时频图特征结合卷积神经网络在不同测试集可达到(98...     

基于均值特征和改进深度神经网络的说话人识别算法

为提高神经网络在说话人识别应用中的识别性能,提出基于高斯增值矩阵特征和改进深度卷积神经网络的说话人识别算法.算法首先通过最大后验概率提取基于梅尔频率倒谱系数(Mel Frequency Cepstrum Coefficient,MFCC)特征的高斯均值矩阵,并对特征进行噪声适应性补偿,以增强信号的帧间关联和说话人特征信...     

基于卷积神经网络的超声金属材料辨识

传统的金属材料辨识方法会给被检测样品带来一定程度的损伤。文章通过采集金属材料的超声回波时域信号,采用短时傅里叶变换对其进行时频分析,得到包含金属材料微观组织信息的超声时频谱。将目标样品的超声时频谱预处理后作为训练样本,输入到构建好的卷积神经网络中进行训练。然后采集目标样品以及干扰样品的超声频谱图,分别将其输入网络进行辨识。结果表明,神经网络在训练时收敛较快,损失函数在迭代200次后收敛,在经过100次迭代后训练集准确率趋于100%。训练完成的网络模型记录着对应训练样本的特征信息,利用该训练好的网络对待测样本进行辨识,最终可实现超声金属材料辨识。     

DeepESC网络的环境声分类方法研究

为进一步提升环境声分类的识别率,提出了一种仿深度隐藏身份特征（Deep Hidden Identity Feature,DeepID）网络连接方式的卷积神经网络——深度环境声分类网络（Deep Environment Sound Classification,DeepESC）。DeepESC网络共有六层——三层卷积层、两层全连层以及一层聚合层,为使网络在自动抽取高层次特征的同时能有效地兼顾低层次特征,网络将三层卷积层的输出聚合为一层,该层充分包含不同层次的特征,提升了卷积神经网络的特征表达能力。ESC-10和ESC-50数据集上的仿真结果表明：在相同的识别框架下,与随机森林分类器相比,本文网络识别率分别平均提升了7.6%和22.4%,与传统的卷积神经网络相比,识别率分别平均提升4%和2%,仿真实验验证了本文分类器的有效性。