Search

In this paper, we propose a transfer learning-based dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) method for classifying fibroadenoma and invasive ductal carcinoma (IDC) in breast tumors. A total of 207 breast tumors from patients were collected and identified by pathologic diagnosis within 15 days after enhanced DCE-MRI examination; 119 patients (average age 50.52±10.33 years) had fibroadenomas, and 88 patients (average age 42.20±10.10 years) had IDCs. Two lesion-level models were built based on the InceptionV3 and VGG19 models, which were pretrained with the ImageNet dataset. The effects of different depths of transfer learning were examined. The network's performance was evaluated through five-fold cross validation. In the lesion-level models, the model based on Inception V3 obtained better results (area under the receiver operating characteristic curve (AUC) = 0.89) when the weights were frozen before layer-276. The other model based on VGG19 obtained better results (AUC = 0.87) when the weights were frozen before layer-13. Compared with the image-level models, both lesion-level models displayed better discrimination (accuracy increased by 13% and 14%) based on the VGG19 and Inception V3 models, respectively. Our research confirms that transfer learning can be utilized to classify fibroadenomas and IDCs in DCE-MRI images. Different depths of transfer learning result in different performances, and our proposed lesion-level model notably improves the classification accuracy.

The development of wireless capsule endoscopy (WCE) allows painless imaging of the internal digestive environment. Limited by the image acquisition and transmitting time, the frame rate of WCE device is not high enough to generate visually consistent video data. Therefore, in the present contribution, we propose a robust WCE video frame interpolation method, aiming at generating consistent and detail-rich gastrointestinal (GI) track video frames. Considering the special environment of digestive system, we firstly correct the WCE device image capturing position based on stereo geometry to align adjacent frames, and then estimate optical flow using PWC-Net flow estimation architecture. Using the well refined optical flow, intermediate frame between two adjacent frames is interpolated and projected back to the original WCE image form. Comparing with a state-of-the-art method DAIN, our work outperforms on both visual performance and frame prediction accuracy. This work significantly improves the smoothness of WCE video, and provides more completed image data for subsequent data processing.

When connecting to microgrid in traditional distribution network, interconnected system faces a series of problems to be solved from long-scale power flow calculation to short-scale economic operation analysis to ultra-short-scale stable operation. In this paper, aiming at the problem of interconnection system planning in long-scale simulation, the influence of different access points, different penetration rates and different combined power generation conditions on system power flow when microgrid is connected to distribution network is simulated. The simulation results show that the access of microgrid will bring great influence to the power flow of distribution network. Therefore, when the system is running, it should be selected to dispatch the microgrid when its external power is small, so as to avoid excessive fluctuation of the system and affect the power supply quality.

In response to the snake-like deviation phenomenon in the fabric production process, an embedded development system based on ARM Cortex-A9 for fabric edge detection is designed. The system uses an edge detection algorithm based on machine learning, supplemented by image preprocessing and Huff's line detection algorithm. The embedded operating system and the OpenCV computer vision library are ported on the ARM side, a cross-compilation environment is built, and QT- GUI programming for real-time display. Experiments prove that compared with traditional systems, this system has better adaptability and can more effectively track the edges of cloth.

In this paper, we propose an analysis method for defect data of capacitive devices and a method for predicting the occurrence time of defects based on the analysis method. This method is not sensitive to noise in data and has strong robustness. We analyze various features that may affect the occurrence time of capacitive device defects, and use sparse auto encoder (SAE) for dimensionality reduction and denoising. This unsupervised method learns the lowdimensional sparse features of nonlinear manifolds from the data features, which improves the generalization performance of the classifiers to a certain extent. Based on K-nearest regression (KNR), support vector regression (SVR), random forest (RF), gradient boosting decision tree (GBDT) and deep learning (DL) models, we predict the occurrence time of capacitive equipment defects respectively. The root means square error (RMSE) between prediction time and labels is only 243.80 days. Compared with the simple baseline model, the error is reduced by 1773.50 days, and the predicted result is state-of-the-art.

Good data quality is good for later data analysis. Due to sensor failures, communication failures and other issues often lead to abnormalities in the photovoltaic(PV) database. In order to obtain high quality data, data cleaning is essential. The great instability of PV data itself has brought great difficulties to PV data cleaning. In this paper, based on the characteristics of PV data, the data cleaning process and core algorithm are proposed. Firstly, different abnormal data are identified, and then different abnormal data, ie redundant data, default data and error data are deleted, complemented and reconstructed respectively. Among them, a Pearson correlation-based reconstruction method is used for a single set of abnormal data, and a cubic spline interpolation-based reconstruction method is used for data in which both sets are outliers. The Matlab simulation results of real data in the PV database verify that the data cleaning method is effective and better than other commonly used reconstruction methods.

This paper describes a wearable intraoral system for speech therapy using artificial sensory feedback timed with the undesired tongue movement. The system has been implemented as a custom-made palatal retainer, which includes two optical distance sensors and two stimulators to provide error feedback to the tongue. Subjects wore the palatal retainer to test the system performance. By testing the system with phonetic targets /t/ and /d/, we showed that the system was able to detect the tongue movement during pronunciation. The system was also tested to see if it can help a non-native English speaker with Hindi mother tongue to correct the pronunciation of alveolar consonant /t/. The subject was asked to read multiple words containing /t/ consonant repeatedly, with distracting words in between to minimize the involvement of cognition or intentional correction. Test results indicated that the error feedback via stimulation helped the subject to move his tongue forward towards the alveolar ridge in pronouncing /t/. The result justified our hypothesis that intrinsic sensory feedback can be an effective way to train non-native speakers to correct their pronunciation.

Rely on the distributed photovoltaic management system database, this paper puts forward a dynamic optimization of Mysql configuration parameters using hill-climbing algorithm, and through the index on the database field and multi-thread query optimization methods, make every effort to reduce the time spent on querying the data in database, and through a set of simulation, The results show that the parameters tuning is effective, the design scheme using index and multi thread can reduce the query time by more than 45% compared with the traditional single thread query method when querying a large amount of data in the distributed database, so it can achieve the database query optimization, improve the operation efficiency of distributed photovoltaic management system.

Microwaves are indispensable part in the modern kitchen. In the transforming of traditional kitchen to be smart, microwaves are beginning to combine internet with mobile phone APP. This kind of information interconnection changes the control interface of traditional microwaves, which needs more functions to achieve remote operation. However, it's difficult for users to adapt these new interfaces which now increases the difficulty and psychological burden. The purpose of this study is to optimize the information-linkage methods for users, mobile phones, and intelligent microwaves, promote information interconnections and enhance the user's experience. This paper analyzes the similarities and differences between traditional microwaves and intelligent microwaves, optimizes the information linkage processes, and verifies the design results through the emotional evaluation. We selected 40 subjects, tested and evaluated the experience for these new interfaces, and the results of emotional evaluation showed that the new interface prototypes optimized the information linkage and improved the usability. This new design allows more users to accept and be willing to further use intelligent microwaves.

Space target ISAR image components segmentation is an important foundation of space target recognition based on ISAR image. Considering the difference between the ISAR image and optical image and the main body has the diverse distribution of pixels from the subcomponent in the ISAR image, this paper proposes a segmentation idea that subcomponent segments after the main body based on CLEAN: firstly, we develop an adaptive region growing method to segment the main body that emphatically analyses and solves the setting problem of the threshold of region growing method; then, in order to divide the residual subcomponent, we exploit the proximity of pixels in the ISAR image to propose an iterative correlation clustering method based on perceptual organization. Finally, some simulation experiments are conducted to verify the validity of our methods.

Lumped parameter model was used to explain the working principle of magnetically coupled resonant wireless power transmission (MCR-WPT) system. A resonant magnetic coupling two-coil model with Series-Series (SS) structure was established based on the equivalent circuit theory. The mathematical expression of output power and transmission efficiency of the MCR-WPT system was obtained. The relationship between transmission efficiency and system frequency was analyzed, as well as the relationship between transmission efficiency and transmission distance. A magnetically coupled resonant WPT system was built. The experimental results verified the theoretical analysis.

Reliable operation of the power transformer is the basis for the security and stability of power system operation. The traditional state evaluation methods for transformer are inadequate, especially in device properties and environmental information. This paper proposed a transformers environmental and properties state evaluation method based on the defects analysis. Firstly, statistics transformer defect cases in Guangdong. According to the causes, classify the defects as leakage (from tank, oil conservator, and other main parts of transformer), moisture, and component fault, overheat, rust, filthy, exceed the standard. By the statistics, it shows that moisture, leaks, component failure, overheating defects are the main causes. Secondly, analyze the relevant between main causes and influence factors such as operational life, voltage levels, temperature, and rainfall. Then score each factor, according to the environment of the transformer, to evaluate the specific defect's risk. The code of points comes from count general distribution of defect. According to the distribution law of the factors, the basic risk criteria are determined by the total defect's distribution, and the different kinds of influence risk criteria are determined by the distribution characteristics of specific defects. At last, summarize specific defects score to get the corresponding operation and maintenance strategy, which are divided into 4 grades, healthy, potential, preliminary, and Critical. In the end of paper, test the assessment method with a running transformer case. The test result shows that the assessment method successfully evaluates a power transformer's device properties and environmental state.

With maturity of micro-electro-mechanical system (MEMS) technology, varieties of electronic products have entered into the era of integration, miniaturization and low power consumption. It is noted that the traditional battery power supply method restricted the development of electronic products. Vibration energy harvesting is one of the most promising solutions. Because the vibration frequency in normal circumstance is about 1–100Hz, the lower resonant frequency of piezoelectric harvester is demanded. Based on the piezoelectric effect, the characteristics of a novel kind of Double-Fork piezoelectric energy harvester (DFPEH) was studied. A model of the DFPEH was built in the finite element analysis software ANSYS, and the corresponding static analysis, modal analysis, harmonic response analysis were made sequentially. Results show that the first three order resonance frequencies of the harvester are 21Hz, 47Hz and 91Hz separately, which are suitable for low-frequency vibration environment. The frequency gap between the first-order resonant frequency and the second-order resonant frequency is 26Hz. Besides, the 3dB bandwidth of the first-order resonant frequency is 8Hz under condition of acceleration value 9.9m/s2 and the damping factor 0.025, which is applied on the fixed end of the DFPEH. Consequently, the DFPEH has shown the distinguished broadband characteristic.

With the development of intelligent technology, intelligent products start to come into people's sight. Wider thoughts are provided to enhance the efficiency of utilization of water resources in urban families because of the intelligent technology. Based on literature research, we regard the technology of water grading and saving as starting point and technical support. According to analysis of intelligent technology and water-saving scheme, theoretical exploration is conducted into intelligent water saving products and system with idea of increasing income and reducing expenditure in home bathroom. We provide prototype design of intelligent water-saving system and focus on faucet redesign in bathroom based on the concept of Reuse and Reduce respectively. The experience accumulated during the process of designing practice lay the foundation of summarizing principles of household water-saving products and system design. The study discusses the possibility of intelligent technology applying to the water-saving bathroom and points out the specific design principle and future hotpot of smart home bathroom, which is of a practical and theoretical significance for the water resources conservation researches.

With the development of intelligent technology, Smart Home began to integrate into people's family life, and intelligent technology can provide more ideas for the establishment of a healthy home ecology. In order to better solve the security risks of family ecological environment, maintain the health of families, we analyzes the application and development of wisdom life philosophy, family ecological system, existing home products and intelligent technology in family life through the literature research, user research, design practice and design evaluation. At the same time, the content of the study also relates to the composition and maintenance of the family ecosystem, people and Smart Home products in the operation, emotional exchange and innovative presentation form. The research results can be used to control the temperature, humidity and air in the home environment. This method put forward a new idea to promote a friendlier and healthier family life in theory, and it has been put in practice to protect the health of the users in the family environment.

In 2015, the intelligent hardware has entered the rapid development period, and the global intelligent product shipments increased by 164%. Although the hardware of intelligent become a trend, but its biggest problem is that the intelligent hardware is not to build more intelligent life scenario after hardware connected to the Internet, and this problem let the consumers have no feelings about the benefits of intelligent, also it has created the problem of product acceptance remains low. In order to solve this dilemma, the research start from the perspective of creating intelligent life scene, then the research let users really experience the convenience of intelligence life by providing users with multimodal interaction. The research explore the reasonable way of intelligent hardware interaction design by intelligent hardware prototype design and evaluation, then summarizes the design principle of multimodal interaction. The research conclusion proves the multimodal interaction design method can effectively improve the efficiency of the use of intelligent hardware, and increase the users' satisfaction.

Smartwatch as the representative of wearable devices is rapidly universal, but it is different from smart phone and the satisfaction of the people after the purchasing of smartwatch is not high. To solve this realistic problem, research on the reasonability of gesture interaction by taking the promotion of natural interaction as the target, interaction efficiency of smartwatch as research content, and aiming at the process of gesture interaction of smartwatch and using the task evaluation in user's operation. Action process model of gesture interaction of smartwatch is proposed through the experimental research and data analysis. This model helps user to make natural interaction with smartwatch and enhance user experience. This model not only can be used in smartwatch or other wearable devices, also can be used as a reference standard for evaluation of gesture interaction of smartwatch.

It is known that lots of MRI imaging reconstruction approaches usually encounter ill-posed problem. To address this problem, we propose a novel method for MRI imaging reconstruction via a new regularization constraint term. The regularization term is inspired by elastic net and is a combination of elastic net and total-variation(TV). The proposed MRI imaging reconstruction model is solved using the alternating direction method of multipliers(ADMM). In addition, our approach also achieves good performance for image reconstruction under noised image situation. Experimental results show that our method is effective and superior to the traditional methods.

According to the research status of radar, this paper presents a MIMO cross array with azimuth-elevation combined beamforming. Compared with a M×N rectangular array, the number of required array elements could be reduced from M×N to M+N+4 so as to drastically reduce the system cost while maintaining higher angular resolution performance. Compared to a M+N cross array with azimuth-elevation combined beamforming, the proposed MIMO cross array could achieve higher angular resolution and use efficiency of antenna with 4 additional transmitting antennas.

Our previous work has shown that Deep Bottleneck Features (DBF), generated from a well-trained Deep Neural Network (DNN), can provide high performance Language Identification (LID) when Total Variability (TV) modelling is used for a back-end. This may largely be attributed to the powerful capability of the DNN for finding a frame-level representation which is robust to variances caused by different speakers, channels and background noise. However the DBF in the previous work were extracted from a DNN that was trained using a large ASR dataset. Optimal LID DBF parameters may differ from those that are known to be optimal for ASR. Thus this paper focuses on investigating different DBF extractors, input layer window sizes and dimensionality, and bottleneck layer location. Additionally, principal component analysis (PCA) is used to decorrelate the DBF. Experiments, based on the Gaussian Mixture Model-Universal Background Model (GMM-UBM) operating on the NIST LRE 2009 database, are conducted to evaluate the system. Results allow comparison between different DBF extractor parameters, as well as demonstrating that LID based on DBF can significantly outperform the conventional shift delta cepstral (SDC) features.

A low-power energy harvesting system targeting to harvest several microwatts from vibration is presented in this paper. The system consists of a piezoelectric vibration energy harvester and a power condition circuit. The proposed circuit store energy by an integrated bimorph piezoelectric generator that extracts energy from ambient vibrations. The generator is fabricated using thick film printing technology. The experiment measured that the generator produced a maximum RMS output power of 100 uW when excited at vibration with a frequency of 30 Hz and peak amplitude of 0.5 g (4.9 m·s -2 ). Our research show that the proposed system can store the generated power effectively and deliver a smooth and reliable DC to power up a wireless sensor node (WSN).

Contactless power transmission system can supply power in non-contact way for many electronic equipments, it has become a hot topic in recent research. Take topological structure of serial-parallel CPT system for example, the optimal values of the compensation capacitors which can solve the poor coupling problem of the separable transformer are determined. A frequency tracking control system based on single chip microcomputer was designed, which can overcome the problem of system resonant frequency changed in the conditions of load switch and different transmission distances to improve the transmission power. Simulation and experiment proved the correctness of theoretical analysis.

Many successful methods for image classification transform low-level descriptors into mid-level features, and obtain a compact and discriminative representation. However, these approaches rarely think about the spatial context information between mid-level features. In this paper, we present a new mixed-order structural (MOS) representation which may generalize the widely used method spatial pyramid representation (SPR) with correlation modeling method. We first partition the image following SP and extract the mid-level feature for each spatial region. Then our MOS consists of 0th and 1st - order correlation model based on mid-level features. The 0th- order correlation can answer what the mid-level feature represents, and the 1-th order corresponds to spatial contextual information in ambient neighbors. In addition, we apply principal component analysis to reduce the dimension of mid-level features which makes our MOS can be a relatively low dimensionality. Our experiments demonstrate that the proposed MOS representation achieves a better performance compared with spatial pyramid matching (SPM) and other state-of-the-art algorithms.

For the huge amount of the logical path, the path coverage technology of statement level can't be realized. On the premise that each function has finished the unit testing, the path coverage algorithm Based on the function calling relationship can guarantee the adequacy of the test. Get the dynamic path by running the instrumented source code, then split the dynamic path using the splitting algorithm. Match the split dynamic path called subset of the global static path set with the global static path set using the matching algorithm. The coverage rate can be figured out with the matching result. The experiment results show the splitting algorithm and the matching algorithm can obtain the coverage rate accurately.

This paper proposes a novel exemplar-based language recognition method for short duration speech segments. It is known that language identity is a kind of weak information that can be deduced from the speech content. For short duration speech segments, the limited content also leads to a large intra-language variability. To address this issue, we propose a new method. This borrows a vector quantization based representation from image classification methods, and constructs the exemplar space using the popular i-vector representation of short duration speech segments. A mapping function is then defined to build the new representation. To evaluate the effectiveness of our proposed method, we conduct extensive experiments on the NIST LRE2007 dataset. The experimental results demonstrate improved performance for short duration speech segments.

Height estimation of man-made structures in urban and suburban areas is an interesting but challenging problem, especially from single SAR image, which usually occurs in emergency response after crisis. A novel model based height estimation approach is proposed, which finds a best match between parameterized object model and the observed single intensity SAR image data. Firstly, a projection function for single bounce signal return between 3D object model and corresponding characteristic regions in SAR image is constructed, which make the model generation phase very simple and fast. Then, a new likelihood measure for the matching between the height hypothesis and the SAR data is proposed in the respect of image segmentation. The matching procedure is iterated until a global maximum similarity is achieved and then the height estimation is obtained, which is realized by the simulated annealing algorithm. The experiments performed on general terms of buildings in the simulated SAR image data set verify its validity.

With the improvements of spaceborne and airborne SAR system resolution, the applications of radar remote sensing has been extended to building 3D geometric information retrieval and reconstruction from urban SAR images, which is the foundation of build-up areas reconstruction and urban analysis. This paper mainly focuses on the problem of building height estimation from a single high resolution (HR) SAR image of urban scenes. A model based method combined with image segmentation framework of building height estimation is proposed. This method optimizes a new likelihood measure between the projection image from 3D geometric model of the buildings and the observed image over the heights hypothesis space. With assumption of the parallelepiped shapes, the SAR building area is partitioned into several regions. The new likelihood criterion then measures both the inner homogeneity of partitioned regions as well as their inter heterogeneity to achieve robust height hypothesis test. The optimization is done by simulated annealing in order to avoid local optimum. The experimental results performed on simulated SAR image data set valid the proposed method.

The rapid development of ocean observing instruments makes the rapid increase in the amount and types of marine data, but with the deepening of the marine development, comprehensive utilization and research, traditional data management and use methods have been unable to meet all kinds of users. This paper proposed to implement distributed data management and application through service technology based on the analysis of the situation and lack of marine data management, and designed the system framework and application modes. Finally, this design has been realized in China Digital Ocean information construction.

Digital ocean prototype system is a key part of the China digital ocean information basic framework and the basic integration platform for China ocean information management and application. The purpose of marine environment data warehouse is to provide comprehensive and multiple marine information services. This paper proposed distributed marine environment data warehouse construction, and gives solutions to the key issues, i.e. data registration, publishing and uploading; metadata catalog service; analytic themes in marine environment data warehouse; logical model of multidimensional data; and data visualization.

The graduation project is the final link of higher education. Its quality is the embodiment of the teaching quality of a university, is the comprehensive application embodiment of specialty knowledge ability learned by students, and also is the evaluation of university graduate's quality. The current employment situation of college students is worrying. Integration of graduation design and employment is proposed based on the status of graduation design and employment of undergraduates. And specific measures in civil engineering are discussed. The method of team coaching is one of the ways to resolve this principal contradiction, to share high-quality educational resources and to improve thesis quality of undergraduates majoring in science and technology. It can offer ideas for instructors of civil engineering and will be helpful for graduate employment.