Abstract:Spatial information theory is the fundamental theory of radar and other target detection systems. The spatial information is associated with the range, the direction and scattering signal of targets. Parameter estimation theorem is an important part of the spatial information theory. The proof framework of the parameter estimation theorem is proposed in this paper, which consists of three original concepts of spatial information, entropy error and sampling posterior probability estimation, and one theorem. The parameter estimation theorem shows that the entropy error is reachable, otherwise the empirical entropy error of any estimator is no less than the entropy error. The establishment of spatial information theory will play a significant role in promoting the development of systematic theories and methods of radar.

Abstract:Aiming at the problem that the direct sparse representation of the over-complete dictionary on the image cannot effectively remove the effect of high-frequency noise, and the image reconstruction quality after compressed sensing is not high, an adaptive dictionary learning algorithm based on truncated nuclear norm and low rank decomposition is proposed. The algorithm firstly uses the truncated nuclear norm regularization low-rank decomposition model to decompose the low-rank part and sparse part of the image matrix. The low-rank part retains the main information of the image, and the sparse part mainly contains high-frequency noise and some object contour information. Then, the low-rank part of the image is divided into blocks, and the image blocks are classified according to the texture complexity of the image block. Finally, a K-single value decomposition(K-SVD) dictionary learning algorithm is used to train multiple over-complete dictionaries of different sizes for different categories. Simulation results show that the proposed algorithm can perform better sparse representation of the image, while significantly maintaining the consistency of image block features and significantly improving the quality of image reconstruction.

Abstract:For the data clustering problem of incomplete information system, the set pair analysis theory is introduced into k-means clustering. At the same time, to better represent the relationship between the sample and the cluster, a set pair k-means（SPKM） clustering algorithm for incomplete information system is constructed. Firstly, a set pair distance measurement method is proposed according to set pair theory, and the measurement method is applied to the k-means algorithm to obtain the preliminary clustering results. Then, for samples belonging to multiple clusters at the same time, the samples are assigned into the boundary region of the corresponding clusters. And for samples belonging to only one cluster, it is assigned into the positive region or boundary region of the corresponding clusters. The clustering results are expressed by three parts, which are the positive region belonging to the cluster, the boundary region that may belong to the cluster and the negative region which does not belong to the cluster. Finally, six data sets in the UCI database and four contrast algorithms are selected for experimental evaluation. Experimental results show that the SPKM algorithm has good clustering performance in accuracy, F₁ value, Jaccard coefficient, FMI and ARI.

Abstract:The information entropy system serves as a fundamental theory of uncertainty description and approximate reasoning, and it has been introduced into rough sets to implement data analyses and intelligence processing. Classical complementary entropy, conditional-entropy and mutual-information can effectively describe roughness and fuzziness, and their system expansion has application significance. In terms of neighborhood rough sets, neighborhood complementary information measures are extendedly constructed, and their heuristic attribute reduction is investigated. According to analytical simulation and granular replacement, neighborhood complementary entropy, conditional-entropy and mutual-information are defined, and their system equation, double bounds and granulation non-monotonicity are achieved. Based on the neighborhood complementary mutual-information, non-monotonic attribute reduction and its heuristic reduction algorithm are proposed. The validity of property and algorithm is verified by decision tables and data experiments. By virtue of neighborhood expansion, relevant information measures and attribute reduction have application prospects.

Abstract:As for sentence-level emotion analysis, current deep learning methods fail to make full use of emotional language resources such as emotion words, negative words and degree adverbs. A new model is proposed based on bidirectional encoder representations from transformers (BERT) and dual channel attention. One channel based on bi-directional GRU (BiGRU) neural network is responsible for extracting semantic features, while the other based on full connection neural network is responsible for extracting emotional features. At the same time, attention mechanism is introduced into both the channels to better extract key information, and the pre-trained model Bert is used to provide word vectors and thereafter adjust them dynamically according to the context so as to embed real emotional semantic into the model. The final semantic expression is obtained through the fusion of semantic features and emotional features from the two channels. The experimental results show that, compared with other word vector tools, BERT has a better feature extraction ability, while the emotional information channel and the attention mechanism enhance the model’s ability to capture emotional semantics, which significantly improves the performance of emotion classification and its convergence speed and stability as well.

Abstract:A physical frame segmentation method based on convolutional neural network is innovatively proposed in the present study to address the difficulty of threshold selection in the frame segmentation method and the problem of poor universality of the method. Firstly, the digital sequence is transformed into an image by following three steps including matrix construction, data compression and matrix expansion. Then, the convolutional neural network is trained with the existing samples and the trained convolutional neural network is employed to identify the frame length of the unknown protocol. Finally, on the basis of frame length recognition, the initial position of the frame is identified using correlation filtering method. Therefore, each frame can be extracted from the bit stream. The proposed method, which has higher accurate recognition than existing algorithms suggested by simulation results, has significant potential in engineering application.

Abstract:With the singleness of the training sample, the phenomenon of overfitting occurs in the deep neural network when used for specific emitter identification (SEI), which in turn affects the accuracy. In this paper, a deep convolutional neural network (CNN) structure based on PID algorithm is proposed to alleviate the problem. The structure builds a feedback loop between the output layer and the input layer of the traditional CNN, transforms the error rate of output layer into the probability of dividing the training set data by using PID algorithm, and inhibits the overfitting by optimizing the composition of training set data. The average recognition rate of the network reaches 92.59%when the method is applied to the recognition of ultrashort wave radio. The variance of the recognition rate is about 1/3 of that of the traditional algorithm, and the training time is reduced by about 35 min, obviously the performance of this method is better than that of the traditional neural network. Experimental results show that the algorithm can enhance the robustness of the deep network and effectively suppress the overfitting phenomenon.

Abstract:Because of the excessive protein species and number of small molecules, the computational complexity of medicine simulation development is enormous. Molecular docking is an important method to study new medicine, so it is very important to improve the efficiency of molecular docking experimental system. The main purpose of molecular docking is to research the interaction and relationship between protein receptors and ligand small molecules. We set up a Hadoop platform and use Hadoop’s powerful parallel computing ability to dock proteins (1ppe and 1uy6) with a number of small ligands with different numbers through simulation experiments. The corresponding adjustment and optimization of the working process are also carried out. Experimental results show that the system can effectively improve the docking efficiency, and has good stability and convenience.

Abstract:Cardiovascular disease is one of the main causes of human death. Based on the modified residual network, we identify ECG signals and combine the modified residual network with dilated convolution to extract global information as much as possible while keeping local information unchanged in feature extraction. The MIT-BIH arrhythmia data set is trained, validated and tested using K-fold cross validation. In the experiment, firstly, the convolution layer is used to collect the input images. Secondly, the modified network is used to extract the features. Finally, the Softmax classifier is used for classification. In the MIT-BIH arrhythmia database, the proposed model achieves 97.20% accuracy, 92.85% sensitivity, 98.29% specificity, 93.16% accuracy and 93.00% F₁ score without any additional artificial features and data augmentation. This research will provide technical support for the detection and recognition of ECG signals to reduce the workload of professional doctors in medical institutions.

Abstract:The prevalence of atrial fibrillation (AF) increases with the age，and the recurrence rate is very high．It is necessary to propose an accurate and fast algorithm to distinguish AF．Based on the fourth statistical theory，a quantitative method for distinguishing AF from normal sinus rhythm (NSR) is proposed in this paper combining the chaos character and Yin Yang nature of heart system．Firstly，the phase space with embedding dimension of 6 and delay time from 1 to 30 is constructed by using R-R interval data．The probability density function (PDF) graph is obtained in turn．Then the horizontal axis of PDF graph is taken as strength ξ，the cumulative sum of longitudinal axis is taken as distribution function x，and the fourth statistical theory parameter k value is fitted by the corresponding relation of ξ-x．Finally，differential summation with k and the result is defined as Ksd. From the experiment， Ksd=0.3 can be an important parameter to distinguish AF from NSR．This study can not only distinguish AF and NSR by describing Yin and Yang，but also open a new path for exploration of the fourth statistical theory and provide an important evidence for the accurate and rapid detection of AF in the future．

Abstract:In order to solve the problem that the existing electrocardiogram（ECG）denoising methods are difficult to accurately separate the overlapped myoelectricity interference and extract “clean” ECG, this paper proposes a method of using cardiac cycle and empirical mode decomposition to denoise the noisy ECG signal. Firstly, empirical mode decomposition is used to decompose the noisy ECG signal, and then the intrinsic modal function components of the signal are determined to be noise or useful signal by cardiac cycle. Finally, the intrinsic modal function components of the useful signal are reconstructed to be ECG signal. For validating the proposed method, the dynamic simulation model of ECG signal is used to evaluate the denoising effect of the method under different parameters of noise; and three groups of real noisy ECG are constructed by selecting baseline drift signal bw, myoelectricity interference signal ma and ECG105, 107 and 123 in MIT-BIH database, respectively. Both the evaluation and experimental results show that the method can remove the myoelectricity interference and baseline drift in ECG at the same time, and the denoising effect is better than the usual empirical method.

Abstract:Fifteen typical features in time domain, time-frequency domain and non-linear dynamic are extracted from the mechanomyogarphy (MMG) signals in neck muscles. They are divided into five feature sets according to their nature, and part of them are constructed to high-dimension feature vectors before reducing the dimension by principal component analysis (PCA), which are applied in the pattern research for head movements. The MMG of six head movements (forward, backward, swing to left, swing to right, turn to light, turn to right) are classified by adopting three sorts of classifiers, which are support vector machine (SVM), K nearest neighbor (KNN) and linear discriminant analysis (LDA). Experimental results show that selecting the method of combining features in time domain, time-frequency and non-linear dynamic, and adopting SVM as the classifier can improve the classification accuracy up to higher than 80% in each movement, thus acquiring relatively higher rate.

Abstract:Due to the weak electroencephalogram (EEG) signal during the acquisition process, the EEG is mixed with various physiological artifacts, so it is particularly susceptible to electrooculography (EOG) interference caused by eye blinking and eye movement. A method for constructing a blind deconvolution (BD) model based on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) is proposed to achieve EOG artifact separation. Firstly, the CEEMDAN method is used to decompose the EEG signal containing artifacts into several intrinsic mode functions (IMF). Secondly, the modal component is used as the observation signal to send the EEG signal and the EOG artifacts to form a BD model. Finally, the separation of EEG signal and EOG artifacts is realized by constructing the cost function iteratively. To verify the proposed algorithm, the standard Children’s Hospital Boston (CHB) and the Massachusetts Institute of Technology (MIT) (CHB-MIT) scalp EEG database is used for experimental verification. The correlation between the EOG artifact separation data and the original EEG data is analyzed, and the correlation coefficient is 0.82. The results confirm that this method retains most of the original EEG signal components and has a good effect on the separation of EOG artifacts.

Abstract:Aiming at the problem of slow convergence and low accuracy of image registration method in precision medicine, a registration method based on improved brain storm optimization(IBSO) algorithm is proposed. The new registration includes three steps. Firstly, the unregistered images are decomposed into multi-resolution images. Then, the IBSO algorithm is used for global coarse registration of low-resolution images. Finally, the Simplex is utilized to fine registration of high-resolution images. Compared with methods of particle swarm optimization combined with Simplex, differential evolution algorithm combined with Powell, and brain storm optimization combined with Powell, the average running time of the proposed algorithm reduces by 32.89%, 13.91% and 13.66% respectively in the mono-modality registration experiment, in which the maximum error and the average error are minimum too. It also outperforms the above three registration algorithms in multi-modality registration experiments, in which the measures of mutual information (MI), normalized mutual information (NMI), cross cumulative residual entropy (CCRE) and normalization cross-correlation (NCC) are best in all. Experiments show that the proposed algorithm effectively improves the accuracy and speed of medical image registration.

Abstract:Carbon fiber composite core wires can greatly increase transmission capacity of transmission lines. However, many breaks are caused due to the bending resistance and other reasons, which seriously endangers the safety of line operation. In order to realize on-line defect detection for long-distance transmission lines, this paper proposes an automatic defect detection scheme for carbon fiber composite core conductors. Firstly, the X-ray image of carbon fiber composite core conductors is standardized. Then，data consistency is improved to provide conditions for automatic analysis of conductors after bending compensation and brightness normalization. Finally, the deep convolution neural network technology is used for defect detection. Experiments on aluminum conductor composite core (ACCC) show that the scheme can quickly and automatically identify the defects of carbon fiber composite core conductors.

Abstract:Due to the particularity of the aviation industry, the specific mechanical properties of the aircraft parts provided by some external suppliers cannot be obtained, which makes it difficult to calibrate the accuracy of the strain and load monitoring results in the ground and flight tests. In order to meet the requirement of measuring the elastic modulus of solid structures with unknown material properties, this paper presents a measurement method based on the principle of late echo formation in ultrasonic testing. The transverse wave velocity of materials is measured by using the commonly used vertical ultrasonic probe, and the elastic modulus of materials is calculated according to the inherent relationship among the elastic modulus, Poisson’s ratio, density, longitudinal wave velocity and transverse wave velocity. This is a general method to measure the elastic modulus of solid structures with unknown material properties without additional test blocks, transverse wave probes and complex test equipments. Experimental results show that this method can measure the elastic modulus accurately.

Abstract:A data compression algorithm in seismic data acquisition system is implemented to improve the data transmission efficiency of the acquisition system, which improves the accuracy and depth of marine seismic exploration. This compression is aimed at seismic data stream, which is easier to implement on FPGA than the existing seismic data compression algorithms. Considering the physical characteristics of seismic wave and using the speech compression algorithm for reference, a lossless compression algorithm for 24-bit seismic data stream, which is easy to be implemented on FPGA, is implemented. On average, the collected seismic data can be compressed to 54% of the original data size.

Abstract:The detection and maintenance of the railway contact network is very important for the safe operation of the railway. How to detect the current state of the contact network automatically, quickly and accurately has become an urgent problem to be solved by the construction, operation and maintenance departments of the contact network. In this paper, a patrol inspection robot that can walk freely on the contact network and detect the vibration frequency of the contact network is designed. Combined with the microprocessor with 51 single-chip microcomputer as the core and bluetooth wireless communication module, the vibration frequency detection and control system of the contact network is established. In order to verify the feasibility of the system design scheme, the railway contact network is taken as the detection object to complete the detection of its internal vibration frequency. The experimental results show that the vibration frequency of the contact network can be detected successfully by this detection control system, and the internal damage of railway catenary can be preliminarily judged. Compared with other railway contact network detection methods, this detection and control system is simple and versatile, and it has certain reference value.

Abstract:To realize the real-time switching of the storage rate of the Artix-7 series acquisition platform, an eMMC array controller with independent configuration interface is proposed. The user can not only improve the design efficiency of the storage control logic, but also implement multiple backup solutions based on multiple operating frequencies and control modes of the eMMC array by using the configuration interface. Compared with other similar controllers, it has higher integration reliability and greater application flexibility. The operating principles and technical key points of open-ending multi-block writing, multi-block reading, and erasing in the eMMC 5.1 protocol are firstly introduced. Then the overall design and the control flow of the main modules. Finally, the self-designed test module is used to complete the rate switching function test and the reading and writing rate performance test on the hardware platform. Test results show that the controller can work stably for a long time and achieve the rate switching between 200 MB/s and 800 MB/s under the operating frequency of 200 MHz, meanwhile the maximum rate of each eMMC can reach 200 MB/s.

Abstract:To ensure the safety of blind and visually impaired patients and reduce false alarms in the process of obstacle avoidance, a walking auxiliary system based on gait detection algorithm is designed. In the signal processing algorithm, wavelet packet is used to reduce the noise of the signal, the mean value, standard deviation, variance and wavelet energy of the three-axis acceleration are extracted, and multi-dimensional parameters are constituted as the gait characteristics by combining the information of foot pressure. Then, the optimized multi-classification support vector machine algorithm based on genetic algorithm is selected to train and identify the gait. The hardware part mainly consists of three-axis acceleration signal and plantar pressure signal acquisition hardware design. The software part mainly includes the acquisition and processing of the plantar pressure signal and three-axis acceleration signals and an android phone APP which can dial automatically. Experimental results show that the proposed system is based on the gait detection technology of multi-sensor fusion, with an average gait recognition rate of 90.48%. The system has good portability, low power consumption, good obstacle detection effect, and has certain research significance and practical value in the field of auxiliary walking.