Use of Local Region Maps on Convolutional LSTM for Single-Image HDR Reconstruction

Low dynamic range (LDR) images captured by consumer cameras have a limited luminance range. As the conventional method for generating high dynamic range (HDR) images involves merging multiple-exposure LDR images of the same scene (assuming a stationary scene), we introduce a learning-based model for single-image HDR reconstruction. An input LDR image is sequentially segmented into the local region maps based on the cumulative histogram of the input brightness distribution. Using the local region maps, SParam-Net estimates the parameters of an inverse tone mapping function to generate a pseudo-HDR image. We process the segmented region maps as the input sequences on long short-term memory. Finally, a fast super-resolution convolutional neural network is used for HDR image reconstruction. The proposed method was trained and tested on datasets including HDR-Real, LDR-HDR-pair, and HDR-Eye. The experimental results revealed that HDR images can be generated more reliably than using contemporary end-to-end approaches.     

基于山东省不同模型的物流需求预测比较研究

目的 过对不同预测方法的误差对比研究,选取预测生鲜农产品物流需求量更精准方法,为疫情情况下山东省生鲜农产品市场进行科学性、合理化决策提供参考。方法 公路货物周转量、互联网普及率、GDP、人口数量、第一产业增加值等十大影响因素作为自变量,以生鲜农产品的需求量作为因变量,分别将小波神经网络、人工神经网络(BP)、遗传算法优化神经网络(GA−BP)、粒子群优化神经网络(PSO−BP)、长短时记忆网络(LSTM)等5种方法的数据预测进行对比分析。结果 波神经网络和BP神经网络的预测值明显低于真实值,且平均相对误差接近20%,而优化后的GA−BP、PSO−BP、LSTM算法误差均小于5%,分别为4.06%、1.162%、0.45%,因此,LSTM预测精度最高,效果最好。结论 来山东省的生鲜农产品需求量将持续增长,LSTM算法以其精确度更高,学习能力更强的优点,将会被更多地应用到物流领域研究中。     

基于自身注意力时空特征的语音情感识别算法

针对语音情感识别中无法对关键的时空依赖关系进行建模,导致识别率低的问题,提出一种基于自身注意力(self-attention)时空特征的语音情感识别算法,利用双线性卷积神经网络、长短期记忆网络和多组注意力(multi-head attention)机制去自动学习语音信号的最佳时空表征.首先提取语音信号的对数梅尔(log...     

Classification of Arrhythmia Based on Convolutional Neural Networks and Encoder-Decoder Model

As a common and high-risk type of disease, heart disease seriously threatens people’s health. At the same time, in the era of the Internet of Thing (IoT), smart medical device has strong practical significance for medical workers and patients because of its ability to assist in the diagnosis of diseases. Therefore, the research of real-time diagnosis and classification algorithms for arrhythmia can help to improve the diagnostic efficiency of diseases. In this paper, we design an automatic arrhythmia classification algorithm model based on Convolutional Neural Network (CNN) and Encoder-Decoder model. The model uses Long Short-Term Memory (LSTM) to consider the influence of time series features on classification results. Simultaneously, it is trained and tested by the MIT-BIH arrhythmia database. Besides, Generative Adversarial Networks (GAN) is adopted as a method of data equalization for solving data imbalance problem. The simulation results show that for the inter-patient arrhythmia classification, the hybrid model combining CNN and Encoder-Decoder model has the best classification accuracy, of which the accuracy can reach 94.05%. Especially, it has a better advantage for the classification effect of supraventricular ectopic beats (class S) and fusion beats (class F).     

快递暴力分拣行为视觉识别系统

目的 为实时监测快递分拣过程中粗暴对待包裹的行为,设计一款基于树莓派+EdgeTPU的快递暴力分拣人体行为视觉识别系统.方法 基于TensorFlow深度学习框架,使用PoseNet模型实时采集人体姿态数据,通过LSTM+Attention模型实现人体动作识别,结合MobileSSD进行场景识别,最终实现暴力分拣人体行为视觉识别.结果 实验证明,文中提出的视觉识别方法可以实现暴力分拣5种动作的快速、准确识别,LSTM+Attention人体动作分类模型的测试准确率达到了80％.结论 基于该方法构建的嵌入式暴力分拣行为识别系统,可以实时监测快递分拣中粗暴对待包裹的行为,并实时地告警.     

基于3D和1D多特征融合的语音情感识别算法

针对语音情感识别任务中特征提取单一、分类准确率低等问题,提出一种3D和1D多特征融合的情感识别方法,对特征提取算法进行改进.在3D网络,综合考虑空间特征学习和时间依赖性构造,利用双线性卷积神经网络(Bilinear Convolutional Neural Network,BCNN)提取空间特征,长短期记忆网络(Sho...     

基于AHP与TRIZ的残障人轮椅设计

目的针对轮椅无法满足残障人使用的现状,提出集成层次分析法与发明问题解决理论的残障人轮椅创新设计。方法利用调研法及亲和图法确定残障人轮椅使用需求,并对需求进行层次分级。运用层次分析法获取分级需求重要度,将需求重要度降序排序,准确掌握轮椅设计的重点。建立用户需求—技术方案关系矩阵,验证技术方案是否存在矛盾冲突并运用发明问题解决理论的发明原理寻找最优解。结论依据层次分析法所得客观数据掌握设计重点,通过发明问题解决理论解决技术措施中的冲突,实现了残障人轮椅的创新设计,验证了多理论结合使用的有效性与科学性。     

Ensemble model for the degradation prediction of proton exchange membrane fuel cell stacks

Proton exchange membrane fuel cell (PEMFC) stacks are widely used in mobile and portable applications due to their clean and efficient model of operation. We propose an ensemble model based on a stacked long short-term memory model that combines three machine-learning models, including long short-term memory with attention mechanism, support vector regression, and random forest regression, to improve the degradation prediction of a PEMFC stack. The prediction intervals can be derived using the dropout technique. The proposed model is compared with some existing models using two PEMFC stacks. The results show that the proposed model outperforms the other models in terms of mean absolute percentage error and root mean square error. Regarding the remaining useful life prediction, the proposed model with the sliding window approach can provide better results.     

基于声发射和长短时记忆神经网络的端对端燃气高压调压器故障诊断

在燃气输配系统中燃气调压器处于关键位置,及时有效地诊断其故障类型具有重要意义.文章提出了一种基于声发射信号和长短时记忆神经网络的端到端故障诊断方法,直接利用时域声发射信号对调压器的运行状态进行诊断.首先,设计一个二阶巴特沃斯高通滤波器对采集的声发射信号进行预处理;其次,利用长短时记忆网络(Long Short Term...     

基于模糊c均值聚类算法的控制图模式识别

控制图模式识别能够区分制造过程中的一般因素与异常因素,提高制造过程中的产品质量,减少成本,提高效益。利用蒙特卡洛方法产生样本;采用一维离散小波变换处理原始数据;利用模糊c均值聚类算法进行控制图模式识别。识别准确率99.43%,其标准差为0.002 8。这表明基于该方法的控制图模式识别准确率高,稳定性好,较现有的控制图模式识别方法具有简易、高效等特点。     

Developing a Recognition System for Classifying COVID-19 Using a Convolutional Neural Network Algorithm

The COVID-19 pandemic poses an additional serious public health threat due to little or no pre-existing human immunity, and developing a system to identify COVID-19 in its early stages will save millions of lives. This study applied support vector machine (SVM), k-nearest neighbor (K-NN) and deep learning convolutional neural network (CNN) algorithms to classify and detect COVID-19 using chest X-ray radiographs. To test the proposed system, chest X-ray radiographs and CT images were collected from different standard databases, which contained 95 normal images, 140 COVID-19 images and 10 SARS images. Two scenarios were considered to develop a system for predicting COVID-19. In the first scenario, the Gaussian filter was applied to remove noise from the chest X-ray radiograph images, and then the adaptive region growing technique was used to segment the region of interest from the chest X-ray radiographs. After segmentation, a hybrid feature extraction composed of 2D-DWT and gray level co-occurrence matrix was utilized to extract the features significant for detecting COVID-19. These features were processed using SVM and K-NN. In the second scenario, a CNN transfer model (ResNet 50) was used to detect COVID-19. The system was examined and evaluated through multiclass statistical analysis, and the empirical results of the analysis found significant values of 97.14%, 99.34%, 99.26%, 99.26% and 99.40% for accuracy, specificity, sensitivity, recall and AUC, respectively. Thus, the CNN model showed significant success; it achieved optimal accuracy, effectiveness and robustness for detecting COVID-19.     

GDX1包装机IPC控制系统的研发

目的 为加强银行智能办理业务的设备性能,提高票据数字的识别效率,研究一种改进的算法来获得更高的数字识别效果。方法 根据银行票据的印刷数字特性进行字符的提取和分割,经过图像采集、降噪、二值化之后使用起点直方图法结合步长法进行字符的分割,然后使用改进的LENET卷积神经网络用于提取数字特征,进行分类。结果 通过实验,结果表明文中提出的方法进行复杂环境下的印刷数字识别,准确率达到95%以上,识别速率为1.169 s/张。结论 利用新的字符分割算法与改进的LENET神经网络相结合,可以很好地识别干扰强的印刷票据,准确率高。     

机器人柔顺装配夹具的双小脑模型神经网络控制系统

介绍了一种新型的机器人柔顺装配夹具及其装配系统。设计了由装配力信号为输入的小脑模型神经网络CMAC1和以浮动平台位移信号为输入的小脑模型神经网络CMAC2并行叠加构成的双CMAC神经网络控制器，前者作为前馈控制，后者当夹具刚启动时为前馈控制，装配开始后转为反馈控制。双CMAC神经网络控制器具有存储容量少、系统较稳定、能减少系统误差影响、装配力小等优点。     

Analysis and Process of Music Signals to Generate TwoDimensional Tabular Data and a New Music

The processing of sound signals is significantly improved recently. Technique for sound signal processing focusing on music beyond speech area is getting attention due to the development of deep learning techniques. This study is for analysis and process of music signals to generate tow-dimensional tabular data and a new music. For analysis and process part, we represented normalized waveforms for each of input data via frequency domain signals. Then we looked into shorted segment to see the difference wave pattern for different singers. Fourier transform is applied to get spectrogram of the music signals. Filterbank is applied to represent the spectrogram based on the human ear instead of the distance on the frequency dimension, and the final spectrogram has been plotted by Mel scale. For generating part, we created two-dimensional tabular data for data manipulation. With the 2D data, any kind of analysis can be done since it has digit values for the music signals. Then, we generated a new music by applying LSTM toward the song audience preferred more. As the result, it has been proved that the created music showed the similar waveforms with the original music. This study made a step forward for music signal processing. If this study expands further, it can find the pattern that listeners like so music can be generated within favorite singer’s voice in the way that the listener prefers.     

基于长短期记忆网络的风电机组齿轮箱故障预测

针对风电齿轮箱状态监测数据的多变量动态时空关联性特点,提出了一种基于长短期记忆(long short-term memory,LSTM)网络的齿轮箱故障预测方法,主要包括离线建模和在线监测两个阶段。首先,以齿轮箱油温为目标预测变量,充分考虑其与其它相关输入变量之间在时空维度上的重要关联信息,对历史监测数据进行训练学习,建立齿轮箱正常运行时的油温监测LSTM模型,通过对预测残差进行评估计算设定相应的检测阈值;然后,将训练好的油温监测LSTM模型用于在线测试,通过模型残差分析和阈值比较实现齿轮箱故障状态的检测和预测;最后,通过风电场测试数据对所提出的方法进行验证。结果表明,相比于其它传统方法,该方法表现出更好的预测性能,能够较早预测故障的发生。     

Transfer learning based deep learning model and control chart for bearing useful life prediction

The remaining useful life (RUL) of the machine is one of the key information for predictive maintenance. If there is a lack of predictive maintenance strategy, it will increase the maintenance and breakdown costs of the machine. We apply transfer learning techniques to develop a new method that predicts the RUL of target data using degradation trends learned from complete bearing test data called source data. The training length of the model plays a crucial role in RUL prediction. First, the exponentially weighted moving average (EWMA) chart is used to identify the abnormal points of the bearing to determine the starting point of the model's training. Secondly, we propose transfer learning based on a bidirectional long and short-term memory with attention mechanism (BiLSTMAM) model to estimate the RUL of the ball bearing. At the same time, the public data set is used to compare the estimation effect of the BiLSTMAM model with some published models. The BiLSTMAM model with the EWMA chart can achieve a score of 0.6702 for 11 target bearings. The accuracy of the RUL estimation ensures a reliable maintenance strategy to reduce unpredictable failures.     

Detection and Classification of Diabetic Retinopathy Using DCNN and BSN Models

Diabetes is associated with many complications that could lead to death. Diabetic retinopathy, a complication of diabetes, is difficult to diagnose and may lead to vision loss. Visual identification of micro features in fundus images for the diagnosis of DR is a complex and challenging task for clinicians. Because clinical testing involves complex procedures and is time-consuming, an automated system would help ophthalmologists to detect DR and administer treatment in a timely manner so that blindness can be avoided. Previous research works have focused on image processing algorithms, or neural networks, or signal processing techniques alone to detect diabetic retinopathy. Therefore, we aimed to develop a novel integrated approach to increase the accuracy of detection. This approach utilized both convolutional neural networks and signal processing techniques. In this proposed method, the biological electro retinogram (ERG) sensor network (BSN) and deep convolution neural network (DCNN) were developed to detect and classify DR. In the BSN system, electrodes were used to record ERG signal, which was pre-processed to be noise-free. Processing was performed in the frequency domain by the application of fast Fourier transform (FFT) and mel frequency cepstral coefficients (MFCCs) were extracted. Artificial neural network (ANN) classifier was used to classify the signals of eyes with DR and normal eye. Additionally, fundus images were captured using a fundus camera, and these were used as the input for DCNN-based analysis. The DCNN consisted of many layers to facilitate the extraction of features and classification of fundus images into normal images, non-proliferative DR (NPDR) or early-stage DR images, and proliferative DR (PDR) or advanced-stage DR images. Furthermore, it classified NPDR according to microaneurysms, hemorrhages, cotton wool spots, and exudates, and the presence of new blood vessels indicated PDR. The accuracy, sensitivity, and specificity of the ANN classifier were found to be 94%, 95%, and 93%, respectively. Both the accuracy rate and sensitivity rate of the DCNN classifier was 96.5% for the images acquired from various hospitals as well as databases. A comparison between the accuracy rates of BSN and DCNN approaches showed that DCNN with fundus images decreased the error rate to 4%.     

基于烟花算法与长短时记忆网络的模温预测方法研究

针对传统神经网络模型预测模具温度准确度低,网络超参数选取困难等问题,提出一种基于烟花算法优化长短时记忆网络的模温预测模型,为铸造成型模温自动控制提供基础。首先根据铸造过程生产工艺选取影响铸造系统的主要变量,利用灰关联分析得出各变量灰色关联度并去除关联度小的变量,建立模具温度影响因子变量的数据集;其次采用烟花算法对长短时记忆网络进行优化,建立模具温度预测模型;最后与BP神经网络和长短时记忆网络预测效果进行对比。实验结果表明基于烟花算法优化的长短时记忆网络的模温预测方法绝对误差小于2. 4℃,平均绝对百分比误差小于0. 12。     

Remaining useful life prediction via a variational autoencoder and a time-window-based sequence neural network

The prediction of remaining useful life (RUL) has attracted much attention, and it is also a key section for predictive maintenance. In this study, a novel hybrid deep learning framework is proposed for RUL prediction, where a variational autoencoder (VAE) and time-window-based sequence neural network (twSNN) are integrated. Among it, VAE is used to extract the hidden and low-dimensional features from the raw sensor data, and a loss function is designed to extract useful data features; by using a sliding time window, twSNN can predict RUL dynamically; meanwhile, it can simplify the network architecture in the time dimension. Furthermore, to achieve higher performance on various failure conditions, long short-term memory (LSTM) cell and convolutional LSTM (ConvLSTM) cell are designed for twSNN respectively. A case study is completed with a dataset of aircraft turbine engines. It is found that the proposed frameworks with LSTM cell and ConvLSTM cell have better performance on both single failure mode and multiple failure modes. The results also show that the prediction accuracy is averagely improved by 6.65% for single failure mode and 15.05% for multiple failure modes respectively.