Abstract:Android system has been one of main targets of hacking. Since its realease, it has been facing security risks such as root, image tampering, and malicious programs. System security usually overlooks framework layer which may cause high security risks. This study analyzes the component in Android framework and how it works. We present an Integrity Measurement Method for Android Framework (FIMM) to ensure Android framework's code integrity and runtime integrity. We focus on the issue of lack of protection for Android framework. The method is able to measure the integrity of component of Android framework and verity it. We consider that Android system service has a long running period, then we analysis the calling process for Android system services and present a dynamic integrity measurement for process providing fine-grained measurement. It measures and verities system service process every time system service is called. Finally, the FIMM impelemention based on Android emulator is discussed and presented. We believe that FIMM achieves the security goals, nevertheless, it causes a little performance loss as well.

Abstract:The bird strike is one of the most important reasons that threaten aviation safety and cause aviation accidents. The National Civil Aviation Administration stipulates that bird ecological surveys and bird strike risk assessments must be carried out for new airports. Bird strike risk in three-dimensional (3D) realworld is an abstract concept. The visualizing bird strike risk can make the airport bird strike prevention work more targeted. This study proposed a 3D visualization method of bird strike risk based on game engine. Taking Xiamen Xiang'an International Airport as an example, we developed a bird strike risk visualization system to realize this method. Firstly, Digital Elevation Model (DEM) data and remote sensing images around the airport were used to construct 3D terrain model. Next, the airport clearance map was used with 3DMAX to build 3D model. Then, the bird strike risk was assessed according to landuse, birdwatching data, and other groundtruth with geographical information system software as a tool. These 3D scalar field of bird strike risk and bird flight path were simulated on Unity3D, a serious game platform. Finally, functions such as fusing all the above elements together, displaying the risk scenarios dynamicly through camera roaming and aircraft flight simulation were developed. These functions can better show the spatial distribution and level of bird strike risk around the airport, and thus can carry out more targeted bird strike prevention strategies and provide decision support for the managers.

Abstract:Aiming at the problems of DTW algorithm in gesture authentication, the dynamic gesture authentication method for smartphones based on CM-DTW (Constraints Multi-dimensional Dynamic Time Wrapping) is proposed. The method uses smartphone built-in sensors to obtain gesture data which representing users' biological behavior characteristics. The candidate template set of legitimate users is selected by DTW algorithm with Sakoe-Chiba window constraints, and a standard template is obtained by normalizing the candidate template with linear up-down sampling. Compared with DTW algorithm, the proposed method not only improves the time efficiency of user identity authentica-tion but also maintains the accuracy.

Abstract:Emotion is the most important semantic information of music. Music emotional classification is widely used in music retrieval, music recommendation, and music therapy. Traditional music emotional classification is mostly based on audio. Nevertheless, based on the current technology level, it is difficult to extract semantic-related audio features from audio. There is some emotional information in the Lyric text, and music emotional classification is carried out with lyrics. This study focuses on Chinese lyrics and constructs a reasonable music emotion dictionary, which is the premise and foundation of lyric emotion analysis. Therefore, a Chinese emotion dictionary in music field is constructed based on Word2Vec, and Chinese music emotion analysis is carried out based on the weighting of emotional words and the part of speech. Firstly, this study constructs the emotional lyrics table based on the VA emotional model and adopts Word2Vec. The idea of word similarity calculation in Word2Vec extends the emotional vocabulary and constructs a Chinese music emotional dictionary, which contains the emotional categories and emotional weights of each word. Then, according to the dictionary, emotional words weights are obtained, and feature vectors of lyric texts based on TF-IDF (Term Frequency-Inverse Document Frequency) and lexical features are constructed. Finally, music emotional classification is realized. The constructed music emotion dictionary is more suitable for music field, and the accuracy can be improved by considering the influence of part of speech when constructing feature vectors.

Abstract:In recent years, deep learning methods have been widely used in the financial field, especially promoting the development of stock price prediction. Aiming at the problem of poor accuracy and robustness of univariate Long Short Term Memory (LSTM) network in general stock price prediction, this paper introduces the idea of multiple-factor model in quantitative stock option strategy in the field of economics into stock price prediction, calculates the multiple-factor of stock and takes it as the input feature for the prediction model. At the same time, in order to make the model accept the multiple-factor input, a multi-variable LSTM model for stock price prediction is produced on the basis of the univariate LSTM model. The experiment results show that, with the introduction of multiple-factor model, it not only improves the accuracy of stock price prediction, but also brings better model robustness to some extent.

Abstract:In this study, an effective method about image edge detection is proposed from the local energy. In the symmetrical region centered on one pixel, we compare the differences of the gray value between every pixel all around the region and the pixel in the center. And the local energy of the central pixel is the sum of all the difference squares. The local energy can be used to detect the image edge effectively, because the local energy of the edge point is bigger than the pixel in the smooth region. And edge point can be precisely found, according to the local energy function constructed in the study. The Baddeley Error Metric (BEM) method is used to evaluate the accuracy of the proposed method. The experimental results show that the proposed method is better.

Abstract:Aiming at the problem that the traditional power network traffic detection and security warning system cannot meet the demand in terms of accuracy, timeliness, expansibility, and efficiency in facing of massive high-dimensional data, a Spark based traffic anomaly detection platform for power grid industrial control system is established. The platform takes Spark as its computing framework, which is mainly composed of data acquisition and network traffic deep packet detection protocol parsing module, real-time computing data analysis and processing module, security warning and prediction module, and data storage module, to complete process for traffic anomaly detection. Experimental results show that the platform can effectively detect the abnormal flow, make the safety warning, convenient for staff to make decisions in time. This fully shows that the platform is very suitable for electric control system, can deal with massive amounts of high-dimensional complex data real time analysis and early warning, greatly improve the safety performance of the power grid control system.

Abstract:Aiming at the reusability and practicability of China's important product traceability system, this study analyses the root causes of the problems of current traceability system model construction, traceability coding, and traceability basis. To improve the integrity of traceability information, the adaptability of traceability system and the consistency between system, and business, the traceability process model, traceability coding method, traceability domain model, and traceability data integration method are studied. This model has been applied in the planning and design of several important product traceability system construction demonstration projects. The results show that the model has obvious advantages in reusability and practicability.

Abstract:With the increasing application of Jupyter Notebook in the field of data science, the more functional requirements for multi-user management and cluster computing resource scheduling are increasing. This study, starting with the basic concepts of Jupyter, expounds the influence of Jupyter on the broadcast of scientific research achievements, summarizes the present situation of the research organizations and institutions of higher learning and other organizations in the field of research on the distributed architecture of Jupyter, analyzes the characteristics of the architecture of the Jupyter system in detail, and reconstructs the Jupyter by means of microservice. Through the resource scheduling and allocation algorithm of Kubernetes, a highly elastic distributed architecture based on container technology is implemented. Finally, the test results show that the architecture proposed in this study has been improved to a certain extent on the performance of the access load, and the target of load balancing on the cluster is achieved in the number of users' running.

Abstract:In this study, we mainly focus on the application of deep learning in the classification of antimicrobial drug using methods and data mining. In the process of text data mining using existing methods of using antimicrobial drugs in disease and electronic medical records, we use the Long Short-Term Memory model (LSTM) based on attention model to train the data of antimicrobial drugs corpus, and express and understand the problems by means of self-learning features, so as to avoid the error of extracting artificial features. The maximum classification accuracy is increased to 89.97% compared with the traditional data mining method. As a result, it provides better antimicrobial treatment plans for patients with different diseases. According to the experimental results, the proposed method can automatically learn and generate treatment knowledge base without the need for manual rules, so as to provide the decision-making support for doctors to treat patients.

Abstract:The industrial Internet data management platform is a convenient and efficient management and analysis of industrial data flowing through the network. But because it provides a convenient data management channel, it has a higher risk of data leakage and more strict security requirements. In response to this problem, this study designs a set of industrial Internet data management platform based on China Mobile OneNET platform, which protects data information through classification protection and hierarchical restrictions on data interaction, minimizes the risk of disclosure and destruction of data information while meeting management needs, and improves the security of industrial Internet data information to a certain degree.

Abstract:In view of the negative effect of harmonics on the safety, the economics, and the operation of electric power system, this work proposes an electric power big data based harmonic simulation system which is studied from the aspects of visual modeling, harmonic computation, and the combination of electric power big data and harmonic simulation system. Based on GEF and JavaFX, the grid network is visually built and the simulation results are graphically represented. By extending openDSS, traditional and new harmonic simulation models are implemented. Meanwhile, electric power big data is used as input, which can not only reduce the workload and error rate of manually inputting data, but also get better simulation results. Case study shows that the proposed system is able to support harmonic simulation, is more intuitive, is easy to be operated and extended, and is a new attempt to combing electric power big data and harmonic simulation system.

Abstract:In the information and intelligence age, we urgently need more powerful and flexible industrial monitoring. The self-developed configuration software can not only improve the operational efficiency of the terminal monitoring system, but also can meet the customized needs of the ZPMC and enhance the core competitiveness of the company's software development. This work studies the overall design architecture, key implementation technology and implementation effects of the ZPMC new generation configuration software. The configuration software based on new architecture has a good systematic interactive interface design. It realizes centralized deployment by using the virtualized network and mobile applications, easily grasp the information of production operation and equipment status anywhere at any time. The pioneering technology of multi-screen mapping realizes the dynamic configuration and can display the key content of dock monitoring in a panoramic view with one computer. The configuration software based on the new architecture can significantly improve the efficiency of the real-time monitoring system and optimize the user experience. It has very important reference value for the development and design of domestic configuration software.

Abstract:With the birth of Loognson, Phytium, and SW domestic processor, a number of domestic operating systems supporting domestic processors have emerged, such as JARI-Works, Neokylin, and so on. The development of debugging tools matched with domestic soft and hard platforms is lagging behind, restricting the efficiency of software debugging under the domestic platform. We design an intelligent debugging project aiming at domestic software and hardware platform, integrate tool chains and intelligent tracking and debugging plug-ins for domestic platforms, then we can auto start and intelligently track debugging process, to realize the intelligent debugging in graphic. We build a UI-friendly, easy to use, and self-control integrated development environment. Experiment shows that the project can simplify the debugging flows and improve software debugging efficiency on domestic platform.

Abstract:With the improvement of living standards and the development of the Internet of Things, the demand for intelligent home is becoming more and more urgent. This paper describes the design and implementation of smart home based on raspberry pie. Through raspberry pie as the main module, we build a smart home system to meet the needs of the public. This system takes raspberry pie as the main development platform and develops a smart home solution based on Ubuntu operating system, it contains a speech synthesis, speech recognition, image recognition, data acquisition, AI dialogue, video monitoring, voice control, voice logging, etc. Interact with robots and sensors through voice, WEB, WeChat, and APP. It can interact with robots and sensors through voice and mobile WeChat, and can log in the Web interface to view the underlying data and control the sensors accordingly. The system sensor part adopts ZigBee communication protocol, and MQTT communication protocol is adopted for communication with the server. The two communication protocols are of low cost, low power consumption, and saving network resources.

Abstract:The securities market is an important part of the complete market system, which has an important impact on the operation of the whole economy. The safe and stable operation of the securities trading system is a subject of great concern to the securities industry. This study designs a real-time monitoring system (RSCMS), which can quickly and flexibly monitor the network connectivity and operation of business functions of various trading systems in securities companies, in order to improve the real-time monitoring efficiency of the departments and provide the basis for data evaluation of the operation of the system. This study focuses on the solution of key technology of monitoring system, and provides a successful case in the operation of securities companies.

Abstract:Imbalanced dataset tends to be biased towards "majority" when classifying, and samples generated by traditional over-sampling cannot well express the distribution characteristics of the original dataset. The improved variational autoencoders combine with data preprocessing method, generate samples by the generator of variational autoencoders trained by the minority class samples to balance the training data set, solve the overfitting problem caused by imbalanced dataset of traditional sampling. Experiments are carried out on four commonly used UCI datasets, the results demonstrate that the proposed method shows better classification performance in F_measure and G_mean with high accuracy.

Abstract:For the problem of color deviation caused by mixing near-infrared light into the images taken in the visible light range, the traditional monitoring system uses the mechanically switched infrared cut-off filter to achieve the switching of day and night shooting mode, which is prone to produce mechanical faults and affect the imaging. On the basis of not changing the original CCD or CMOS sensor, this study adopts visible light and 850 nm dual-peak filter to replace the traditional mechanical switching filter, but there is still the 850 nm infrared crosstalk problem. In order to solve the infrared interference, this study abandons the traditional color deviation correction analysis method of infrared interference image, starts with the analysis of the influence of near-infrared light on the sensor, and corrects the spectral characteristics of the camera by ridge regression color calibration method. This process simulates the state of using the infrared cut-off filter when the camera takes the three-band image. First, in the dark environment of the laboratory, use the D65 standard light source box, and use the 850 nm near-infrared light source to directly shoot the camera that has removed the infrared cut-off filter, and shoot the four-band image of the Pantone color card in the shed (RGB three-band and IR near Infrared band); then close the near-infrared light source, taking the three-band image of the Pantone color card with the same position. Color calibration is performed according to the ridge regression algorithm, and a correction matrix between the three-band image and the four-band image is obtained, which is used for color correction of the four-band image to obtain a natural color image.

Abstract:In order to solve the problems of embarrassing sales under the impact of e-commerce and time consuming and exhausting effort of users when chosing the commodity under the "information explosion" of the Internet, this paper introduces the location model of the business-circle based on the circular filtering method and the improved partition-based DBSCAN density clustering algorithm for Zhejiang Province. The geographic location characteristics of the 250 000 merchants' order data in a certain industry of Zhejiang Province are analyzed, and the traditional recommendation algorithm is improved by combining the time decay parameters. The commodity recommendation algorithm for the popularity of the business-circle and the collaborative filtering algorithm for the similarity of the business-circle are proposed. The experimental results show that the algorithm is superior to the traditional recommendation algorithm in terms of recommendation accuracy rate, and to some extent, it alleviates the problem of insufficient cold start and recommended product surprise, which has its practical value and research significance.

Abstract:With the increasing contradiction between energy supply and rapid economic development, building energy conservation has become a key link in sustainable development strategy. It is an important prerequisites for optimal control of building energy conservation that fast and accurate method research for predicting building electricity consumption. In this study, genetic algorithms and ant colony clustering algorithms are combined to improve the network node of IHCMAC (Improvement Hyperball CMAC) neural network based on clustering. As a building power load forecasting model, GIHCMAC (Genetic Algorithm Ant Colony Clustering Algorithm based on IHCMAC) is used to predict the electrical load of an office building in Weifang. The research results show that the prediction model has the smallest number of iterations and high accuracy. Its iteration number, training error, and generalization error are 9, 0.0045, and 0.0014 respectively. Compared with IHCMAC, KHCMAC (K-means Hyperball CMAC) and IKHCMAC (Improvement K-means Hyperball CMAC) model, GIHCMAC has faster convergence speed, higher accuracy, and better generalization.

Abstract:As the static data table in the information system has a low efficiency in development and cannot be changed according to customized requirements, this study proposes a dynamic query technology based on XML configurations. First of all, digester reads the instantiated configuration information from XML which will be handled by the dynamic query engine. Secondly, the dynamic query engine combined with XML configuration obtains the dataset by SQL, HQL, or interface. Finally, the configuration and dataset are transmitted to clients in the form of JSON and shown by improved dhtmlxGrid. After the above steps, the dynamic data table is formed. The technology is applied to the scientific research management system and human resource information system. The application results prove that this technology significantly improves the development efficiency and meets all kinds of customized requirements such as combination-conditions query customization, data column customization, data column rendering, and data range control.

Abstract:With the rapid development of the Internet, the Weibo has gradually become an important way of information dissemination and information collection in social communication, and Weibo retweeting is an important way to spread information on Weibo.The study of the Weibo retweeting problem has a very important significance to Weibo communication, Weibo marketing, and public opinion monitoring. The main factors affecting the retweeting of Weibo are similarity between followers' interest and Weibo text, and changes in Weibo marketing strategy and number of user followers. The previous forecasting models did not consider these two factors comprehensively. To solve the above mentioned problem, this study proposes a method based on recurrent neural network to predict magnitude of Weibo retweeting. First, the SIM-LSTM model is used to build the trend of Weibo retweeting. Then, TF-IDF is used to build the similarity between followers' interest and Weibo text. And finally, neural network model is used to predict whether followers will forward the Weibo. the experiments show that the F1 evaluation value using the proposed algorithm is increased by 5% comparing with other traditional prediction methods.

Abstract:Since the traditional K-nearest neighbor classifier possesses large time and space complexity for larger-scale data sets, prototype selection is an effective processed method which selects representative prototypes (instances) from the original data set for K-nearest neighbor classifier without reducing the classification accuracy. At present, there exist many prototype selection methods. In this paper, based on the existing CURE algorithm, which is difficult to determine the noise points and has bad dispersed of representative points, the shared neighbor density metric is presented to delete noise points and the maximum and minimum distances are employed to obtain scattered representative points, which generates a novel prototype selection methods PSCURE (improved Prototype Selection algorithm based on CURE algorithm). Some numerical experiments are further conducted to show the performance of the proposed prototype selection algorithm compared with other related prototype selection algorithms. The experimental results show that the proposed algorithm not only can select fewer prototypes but also can achieve higher classifier accuracy for almost all the data sets.

Abstract:With the rapid growth in the number of clothing pictures on the Internet, the demand for classification of a large number of clothing is increasing. The traditional use of manual semantic attribute annotation of clothing images does not fully express the rich information in the clothing image, and the traditional hand-designed features can no longer meet the requirements of real precision and speed. In recent years, deep learning has been applied to all aspects of computer vision, laying a solid foundation for clothing classification and recognition technology based on deep learning. In this study, three new sub-datasets are constructed according to the existing dataset deepfashion, the deepfashionkid dataset for classification training, the deepfashionVoc dataset for training with Faster R-CNN, and the deepfashionMask dataset for Mask R-CNN training. The clothNet model is pre-trained on the VGG16 using the deepfashionkid dataset to obtain the clothNet model, which in turn improves the loss function of the Faster R-CNN. And each compares the difference between the two algorithms using clothNet pre-trained model and not used. In addition, this study adopts a new pre-training strategy to adopt a training method similar to grafting learning. Experiments show that these training techniques are helpful for improving the detection accuracy.

Abstract:Aiming at the scheduling optimization problem of hybrid energy storage microgrid, a multi-objective optimization model with economic benefit and pollution treatment cost under grid-connected state is established. Based on the basic fireworks algorithm and the grey entropy parallel analysis theory, a multi-objective grey entropy fireworks algorithm is proposed. The proposed algorithm can effectively handle the conflict relationship between different objectives by assigning different entropy weights to the two studied objectives. The grey entropy parallel correlation degree is adopted as the fitness of fireworks algorithm to select excellent individuals and guide the algorithm to better search region. Simulation results show that the performance of the proposed multi-objective grey entropy fireworks algorithm is significantly better than that of the random weight-based and Pareto-based fireworks algorithm, and better than that of the classical NSGA-Ⅱ algorithm, which verifies the effectiveness of the established multi-objective model and proposed multi-objective algorithm.

Abstract:The classification of polarimetric SAR images by computer has become a research hotspot in remote sensing. In this study, the fully polarimetric SAR data is used to extract characteristics by different algorithms, and the classification of tidal flat of Jiangsu coastal is realized. Firstly, the polarimetric scattering characteristics are extracted by H/α and Freeman decompositions, and the texture features are extracted by gray level co-occurrence matrix. Then, all the extracted features are combined to form different feature sets. Finally, the random forest model is used to classify and accurately evaluate with different feature sets. The study shows that using only texture features to classification achieves a poor performance. The classifications using the scattering features extracted by polarimetric decompositions are better than that of matrix element features. The combination of polarimetric scattering and texture characteristics can obtain best classification in coastal tidal flat, and the overall accuracy and Kappa coefficient are 94.44% and 0.9305, respectively. It indicates that the characteristics of different aspects contained in fully polarimetric SAR image have certain complementarity in the classification of coastal area.

Abstract:Collaborative filtering algorithm is the most widely used recommendation technology in e-commerce system. In order to alleviate the shortcomings of traditional user-based collaborative filtering algorithm in cold start, recommendation accuracy, and data sparsity, this study proposes collaborative filtering recommendation algorithm based on user characteristics. This algorithm extracts the attribute features by using the registration information, extracts the interest features and trust degree from the existing scoring information, and synthesizes the feature similarity of the above features to further generate recommendations. Experimental results show that comparing with the traditional user-based collaborative filtering algorithm, the collaborative filtering algorithm based on user characteristics greatly improves the accuracy of the recommendation.

Abstract:The aim of clinical decision support implementing electronic health records is to satisfy the physicians' information needs. We are motivated to propose an attention based network on query expansion. Considering the difficulty and cost of medical text annotation and inspired by the idea of migration learning, we chose the non-medical dataset for model training, and migrated to medical datasets. The model utilizes LSTM to obtain sentence representation and adopt attention mechanism to obtain entities representation. The proposed approach can dynamically select related entities as expansion of the query. At the same time, we not only consider the score of a single term as an expansion term, but also consider the score of term combination. We conduct the experiments on the three standard datasets of TREC Clinical Decision Support Track, where the approach has a promising overall performance over the strong baseline.

Abstract:Fetal Heart Rate (FHR) is a routine perinatal test and one of the main physiological indices to evaluate the health status of pregnant women and fetuses. In contrast to existing contact FHR detection technology, we propose a convenient and inexpensive noncontact FHR monitoring method based on Eulerian video magnification. The color signal in video is amplified and PhotoPlethysmoGraphy (PPG) is used to extract microvascular blood volume pulsar signal after removing maternal noise. The FHR can be estimated by calculating the power spectral density of the signal. Experiments show that the accuracy of detecting FHR can reach 96% using clinical measured data as the standard.

Abstract:It is popular to develop the city-wide Parking Guidance System (PGS) in China nowadays, in order to alleviate the parking difficulties arising in large cities. Prediction on parking occupancy is the essential intelligent technology to help vehicles find the proper parking lot efficiently in PGS. And the known prediction methods have to be powered by real-time data, without which would cause error accumulation and significant inaccuracies. In the early stage of PGS deployment, however, it is very hard to collect the real-time data from the parking lots all over the city. Therefore, this study takes the historical data of non-stationary parking spaces with periodic characteristics as the research object. Firstly, statistical analysis of parking spaces is carried out according to the central limit theorem and Law of Large Numbers. Then, we propose a method named SAL (non-stationary Stochastic And Long short-term memory) combined with LSTM (Long Short-Term Memory), to predict the parking occupancy at the given time, based on digging the history data. Experimental data prove that compared with using LSTM and Lyapunov exponent method, SAL has lower computational complexity, more accurate prediction, and effectively solves the problem of error accumulation caused by multi-step long-term prediction without real-time data.

Abstract:The highway polysyllabic path problem refers to how to determine a driving path of a vehicle in a highway network with multiple optional paths. At present, the identification point-based polysemy path identification method commonly used in some cases (such as equipment failure, ambient brightness or insufficient transparency) has a low recognition rate, which makes it difficult to identify the vehicle polysemy path in some time periods. Aiming at the above situation, this study proposes a multi-sense path probability identification method based on historical data. The road segment-based clustering method is used to calculate the probability values of each road segment, and then the greedy algorithm is used to find the vehicle's driving path, which is used to identify equipment faults. It assists in identifying polysemy paths. The method can effectively identify the ambiguous path when identifying the equipment failure, and improves the accuracy of the method.

Abstract:Wild wheat is a heterologous hexaploid with a large genome size (about 14 GB) and a lot of repetitive sequences. In order to breed new varieties with good traits, we must first locate the genes that control the target traits. Therefore, it is important to establish a complete and accurate genome annotation process. Traditional genomic annotation method based on database alignment has three obvious disadvantages:first, the alignment runs slowly; second, it is difficult to discover new genes; third, there is no uniform standard for software selection. We propose a new analysis process that combines genetic database alignment, transcriptome high-throughput sequencing, and full-length transcriptome single-molecule sequencing data analysis to annotate hexaploid wheat KN9204 completely and accurately. The annotation of the genome provides an important reference and technical support for revealing the growth of wheat and cultivating new varieties.

Abstract:In this paper, we describe a new feature called Separate Haar (Sep-Haar) feature for fast and accurate face detection. There are three key contributions. "Separate Haar feature" adds a negligible area for the rectangular Haar feature window, by which we can improve the feature extraction efficiency; the corresponding algorithm for selecting the best width of such negligible area is realized by reducing the total number of learned features to reduce the memory used; and experiment result shows that the proposed Sep-Haar feature can achieve best false alarm rate using less number of features in Adaboost algorithm compared with traditional Haar feature. Based on the result, we propose a new classifier that, by using the proposed Sep-Haar features, it can give smaller false alarm rate at each stage, use less number of stages, and at the same time give improved hit rate within the same detection time consummed.

Abstract:The human motion recognition method based on depth learning and depth camera is limited by its application scene, and it cannot recognize the human motion in fast changing scene and static image. This article defines human related angle space, and builds eight four layer BP regression neural network using the human body skeleton recognition based on deep learning framework of data. After data extraction and pretreatment of the bone data of human body, training data is processed to increase the dimension, and then it is fitted through the regression neural network. The experimental results show that the proposed method can effectively regress the human body angle, provide reliable basis for human motion recognition in fast changing scene and static image.

Abstract:Since the selection of the main parameters of the support vector machine can affect the classification performance and effect to a large extent, and the current parameter optimization lacks theoretical guidance, a particle swarm optimization algorithm is proposed to optimize the parameters of the support vector machine. This method improves the shortcomings of the standard particle swarm optimization algorithm with slow convergence rate and easy to fall into local optimum by introducing nonlinear decreasing inertia weight and asynchronous linear variation learning factor strategy. The experimental results show that compared with the standard particle swarm optimization algorithm, the proposed method has good robustness, fast convergence and global search ability in parameter optimization, and has higher classification accuracy and efficiency.

Abstract:Infrared remote control is widely used in wireless communication, and has the advantages of simple protocol, low power consumption and strong adaptability. Based on the communication protocol of infrared remote control, an infrared communication method based on embedded system is proposed. The waveform acquisition, data analysis and decoding work are separated. Compared with the previous method, it has the advantages of complete waveform acquisition, high compatibility, and strong portability.

Abstract:Focusing on the issues that the number, location, and coverage of multi-regional logistics centers of distribution centers are unknown, an improved k-means clustering algorithm is proposed. Based on the urban economic gravity model, this algorithm combines the urban transportation distance with the indicators of household consumption capacity, redefines the distance factor of the similarity measure between objects. The idea of density is introduced into the k-means algorithm, and the concept of intra-class difference mean is raised to determine the optimal number of clusters. After the partition is implemented, the centroid method is used to determine the final distribution center in these areas. Finally, in case study, we analyze the location process of constructing logistics distribution centers in 37 cities in the western region, and compares them with the traditional k-means clustering results. The comparing result shows that the improved algorithm not only saves the delivery time, but also greatly reduces the transportation cost and has sound economic value.

Abstract:The Chinese Weibo is an indispensable communication tool for people today. Mining information in Weibo text is of great significance to automatic question and answer, public opinion analysis and other applied research. The short text classification study is the basis of short text mining. The neural network-based Word2Vec can solve problems of high-dimensional sparseness and semantic gap that traditional text categorization methods cannot solve. This study obtains the word vector based on Word2Vec, then the class factor is introduced into the traditional weight calculation method TF-IDF (Term Frequency-Inverse Document Frequency) to design the word vector weight. Finally, the SVM classifier is used for classification. The effectiveness of the method is verified by experiments on Weibo data.

Abstract:Massive Chinese information processing is a branch of big data processing, and the use of big data technology for Chinese information processing must be inseparable from Chinese word segmentation, so Chinese word segmentation technology is the basic technology of big data Chinese information processing. Chinese word segmentation technology has been advancing in performance and accuracy since this century. In terms of performance, it mainly improves the segmentation scanning algorithm, the word bank storage technology, and query method to improve the performance. In terms of accuracy, it is mainly to improve the processing method of unregistered words and ambiguous words. This paper gives up the idea of searching by lexicon index and proposes a lexicon storage structure based on character tree. Its segmenting speed is 35 times faster than the normal half method, occupying only 1/5 of its memory. It will be a big step forward in the performance of big data technology in processing Chinese information.

Abstract:Temporary discussion groups and the coordinates of the personnel need to be quickly established in such indoor situations as temporary meetings, so as to carry out point-to-point discussion on the purpose of relevant personnel. This study designs and implements instant messaging system based on indoor positioning system. The system is mainly divided into indoor positioning module and instant communication module, and it has realized the functions of group chat in LAN, searching people, and communicating with people nearby. The system is simple in structure and easy to implement. This system is applicable to seminars, product promotion, and other scenarioss. The system has a solid application prospect.