Search

Innovation and development in medical devices are of great significance for overcoming the problems such as the increasing costs of healthcare and the widening societal inequity in medical technologies. This paper presents the design and outcomes of a general education course in medical devices innovation offered for an eight-week quarter each year since Spring 2018 at Zhejiang University. The course consists of two modules, lectures and team project, both of which are well designed based on the entire innovation process spanning needs finding, concept generation, prototyping, and strategy development. A professional teaching team with eight experts from various disciplines and institutes has been established as well. Since its inception, 296 students from 34 majors have participated and 71 original projects have been proposed. The results of self-assessment questionnaires showed that the course had equipped students with broader fundamentals and specialized knowledge, and stronger skills in innovation and teamwork, which provide a solid foundation for students' future innovation practice.

Coronavirus Disease 2019 (COVID-19) has rapidly spread in 2020, emerging a mass of studies for lung infection segmentation from CT images. Though many methods have been proposed for this issue, it is a challenging task because of infections of various size appearing in different lobe zones. To tackle these issues, we propose a Graph-based Pyramid Global Context Reasoning (Graph-PGCR) module, which is capable of modeling long-range dependencies among disjoint infections as well as adapt size variation. We first incorporate graph convolution to exploit long-term contextual information from multiple lobe zones. Different from previous average pooling or maximum object probability, we propose a saliency-aware projection mechanism to pick up infection-related pixels as a set of graph nodes. After graph reasoning, the relation-aware features are reversed back to the original coordinate space for the down-stream tasks. We further construct multiple graphs with different sampling rates to handle the size variation problem. To this end, distinct multi-scale long-range contextual patterns can be captured. Our Graph- PGCR module is plug-and-play, which can be integrated into any architecture to improve its performance. Experiments demonstrated that the proposed method consistently boost the performance of state-of-the-art backbone architectures on both of public and our private COVID-19 datasets.

By driving vehicles in constant spacing, platooning is a promising way to enable a safer and faster mobility with higher lane capacity and energy efficiency. On one hand, recent advances in vehicular communication technologies improve the usefulness of platooning. On the other hand, the string instability in platooning is easily created by the unavoidable communication delays. In this paper, we focus on improving the performance of intraplatoon communications by considering the impact of non-line-of-sight (NLOS), which is typically isolated in most of present works. To do that, we first evaluate the impact of NLOS due to the vehicles as obstacles on signal attenuation among platoon members by developing an analytical mode based on the knife-edge model. To obtain the optimal communication performance, an power control problem is formulated by considering NLOS and multi-user interference. By resorting to the graph neural network (GNN), which can achieve an excellent performance in learning dynamic graph characteristics, an efficient power control policy is developed after modeling the inter-vehicle communication links in the platoon as a fully connected interference graph. Extensive simulations finally validate the performance of our method.

When a gun fires, wear (including erosion) and fatigue cause inevitable damage to the gun barrel. Wear of gun bore gradually results in the decrease of gas pressure, projectile velocity, and accuracy. In contrast, fatigue failure of gun barrel is difficult to detect and measure. In this paper, low cycle fatigue behavior of simulated gun barrel was investigated by using a specially designed apparatus on MTS Landmark. Strain gages were mounted on the outer surface of the tube sample to monitor external strain. Crack propagation was nondestructively inspected by using Zeiss, Xradia 520 Versa. It is found that there is an inherent relationship between external strain and crack propagation. This finding suggests that the fatigue fracture could be prevented by monitoring the external strain on the gun barrel.

Oracle Database In-Memory (DBIM) provides orders of magnitude speedup for analytic queries with its highly compressed, transactionally consistent, memory-optimized Column Store. Customers can use Oracle DBIM for making real-time decisions by analyzing vast amounts of data at blazingly fast speeds. Active Data Guard (ADG) is Oracle’s comprehensive solution for high-availability and disaster recovery for the Oracle Database. Oracle ADG eliminates the high cost of idle redundancy by allowing reporting applications, ad-hoc queries and data extracts to be offloaded to the synchronized, physical Standby database replicated using Oracle ADG. In Oracle 12.2, we extended the DBIM advantage to Oracle ADG architecture. DBIM-on-ADG significantly boosts the performance of analytic, read-only workloads running on the physical Standby database, while the Primary database continues to process high-speed OLTP workloads. Customers can partition their data across the In-Memory Column Stores on the Primary and Standby databases based on access patterns, and reap the benefits of fault-tolerance as well as workload isolation without compromising on critical performance SLAs. In this paper, we explore and address the key challenges involved in building the DBIM-on-ADG infrastructure, including synchronized maintenance of the In-Memory Column Store on the Standby database, with high-speed OLTP activity continuously modifying data on the Primary database.

A method of power transformer mechanical structure fault diagnosis based on controlled switching process vibration measurement is presented in this paper. Compared with steady state loading condition, vibration generated in the controlled switching process contains abundant information of transformer mechanical structure. The wavelet packet decomposition is used to detect the signal envelop of such vibration response signal. Through theoretical analysis, the dynamic characteristic of transformer structure is approximately viewed as a second-order system and by estimating the time constant values, the potential defects of the transformer mechanical structure can be detected.

Frequent pattern mining has been one of the most important data mining tasks, which is widely used in traffic management, animal migration and path prediction. In the view of the large scale, high dimension and real-time updating characteristics of trajectory data in free space, this paper presents an efficient framework for frequent sub-trajectory mining. Firstly, we divide the trajectory twice in order to smooth and reduce the trajectory effectively. Secondly, we expand the distance measure in two dimensions to three to improve the reliability of frequent sub-trajectory mining. Finally, we introduce an incremental clustering algorithm based on density, which greatly reduces the space-time consumption of frequent sub-trajectory mining in real-time updating database. Experimental results demonstrate that our framework improved the efficiency greatly and obtained more reliable results.

In this paper, we propose a Cookie-based Amplification Repression Protocol (CARP) to address the increasing threat of amplification attack. As a replacement of UDP protocol, CARP outperforms previous works in three aspects: i) CARP is a generic solution for all UDP-based amplification attacks regardless of the application protocols; ii) CARP incurs low-latency and introduces no additional latency in most use cases which is suitable for all UDP-based application scenario; iii) CARP supports incremental deployment and plug-and-play. Its interest-driven deployment model makes it much easier to be adopted. We implement a prototype of CARP and evaluate its performance on different metrics. Our results show that CARP is a lightweight and efficient solution for mitigating amplification attack.

DDoS attacks bring huge threaten to network, how to effectively detect DDoS is a hot topic of information security. Currently, there are some methods designed to detect DDoS attacks, but the detection rate of them is low. Moreover, DDoS detection is easily misled by flash crowd traffic. In this paper, a new method to detect DDoS attacks based on RDF-SVM algorithm is proposed. By considering the importance of feature selection in DDoS attacks detection, the RDF-SVM algorithm is designed to exploit random forest to compute the feature importance and SVM to rescreen the features, which will prevent from removing features mistakenly. Finally, an optimal feature subset is obtained, which will reach a higher detection rate and recall rate. In this paper, two kinds of datasets are used to train and test. The experimental result shows that the RDF-SVM algorithm can select the optimal feature subset over KDD99 dataset, and can also distinguish between DDoS attacks traffic and normal traffic (Flash Crowd) over the DDoS dataset collected from real environment. Compared with the CART, Neural Network, Logistic Regression, AdaBoost, and SVM method, RDF-SVM algorithm has a higher detection rate and recall rate.

In this paper the vibration distribution characteristics are introduced with the aim of detecting the potential deformation of power transformer windings. Vibration distribution characteristics are defined based on the vibration measured on the surface of transformer tank. Features of the vibration distribution characteristics graphics are chosen as the indexes that decide whether there are potential winding failures in the transformer. Experiments are taken to validate this algorithm, looseness of compression force and deformation are done to the windings, corresponding data are collected before and after the man-made failures for comparison. From the experiments it can be concluded that vibration distribution characteristics can reflect the health condition of the winds and can be used as a criterion in transformer faults diagnosis.

We propose a coupled system that can produce nonreciprocal channels of light. The system consists of a metal layer sandwiched between two magnetophotonic crystals (MPCs). A resonance mode called Tamm magnetoplasmon polariton (TMP) around the interfaces between the metal and the MPCs can be achieved in this structure. It is found that nonreciprocal light-tunneling channels can be obtained through the coupling of two nonsymmetric TMPs. The nonreciprocal wavelength channels can be adjusted, depending on the incidence direction of the light, whereas the nonreciprocal direction channels can be adjusted, depending on the wavelength of the incident light. Results are demonstrated through electromagnetic-field simulations based on a finite-element solver.

This paper aims to combine the practical application of artificial intelligence and modern navigation collision avoidance, to solve the difficult collision problem of ships which going through the complex water area. Based on AHP (The Analytic Hierarchy Process), ANN (Artificial Neural Network) and other methods, this paper proposes the T minutes prediction algorithm. In this algorithm, the parameter T varies with the changes of ship's specific environment such as the speed, the direction and the weather conditions and so on. According to the value of T, it predict the coordinates of ships T minutes later, marked as P and estimate whether there exist any collision risk there. If the ship is in a dangerous situation, it will sound a warning in advance and build a collision avoidance priority list in accordance with the different risk value of the target ship. This list will be provided to the driver to alert them to take early action to avoid collision according to the different risk level, so as to achieve the purpose of reducing maritime collisions.

Graph coloring is one of the most significant problems in combinatorial optimization. On the basis of the traditional evolutionary heuristic algorithm, this paper presents a memetic algorithm with partial solutions (MAP) for solving this problem. Moreover, we combine a special crossover operator based on the independent sets and a tabu search algorithm with a strong capacity of local search. Meanwhile the score function for individuals and the fitness function for updating process are improved. The experiment for DIMACS Benchmark shows that this algorithm can not only solve the general graphs, but also figure out the optimal solution to the flat series which cannot be solved well by most evolutionary heuristic algorithms. It proves that the memetic algorithm with partial solutions (MAP) has a good stability.

Phase-locking is an important encoding mode in the hippocampal region of brain where neuronal firing concentrates on a certain phase of the periodic field potential, especially θ rhythm. It can be observed during many behaviors while the locked-phase varies accordingly. In this study, the changes of neuronal phase-locking induced by somatosensory stimulation of tail-clamping were investigated. The local field potential and spike signals were recorded in hippocampal CA1 region of anesthetized rats with microelectrode array. The phase-locking relationship was analyzed with Raleigh test and spike-phase histogram. The results showed that during spontaneous activity, 3 of 20 examined neurons were phase-locking to the θ rhythm of the field potential. However, during the tail clamping periods, the number of phase-locking neurons increased from 3 to 8, and the mean locked-phase shifted towards the negative peak of the field potential cycle, which became θ rhythm dominated. These findings suggest that the hippocampus may encode somatosensory information with the phase-locking neurons and by shifting their locked-phases.

Detecting automatic speech recognition (ASR) errors can play an important role for effective human-computer spoken dialogue system, as recognition errors can hinder accurate system understanding of user intents. Our goal is to locate errors in an utterance so that the dialogue manager can pose appropriate clarification questions to the users. We propose two approaches to improve ASR error detection: (1) using recurrent neural network language models to capture longdistance word context within and across previous utterances; (2) using a complementary ASR system. The intuition is that when two complementary ASR systems disagree on a region in an utterance, this region is most likely an error. We train a neural network predictor of errors using a variety of features. We performed experiments on both English and Iraqi Arabic ASR and observed significant improvement in error detection using the proposed methods.

Though sparse features have produced significant gains over traditional dense features in statistical machine translation, careful feature selection and feature engineering are necessary to avoid overfitting in optimizations. However, many sparse features are highly overlapping with each other; that is, they cover the same or similar information of translational equivalence from slightly different points of view, and eventually overfit easily with only very feature training samples in given bilingual stochastic context-free grammar (SCFG) rules. We propose a natural autoencoder that maps all the discrete and overlapping sparse features for each SCFG rule into a continuous vector, so that the information encoded in sparse feature vectors becomes a dense vector that may enjoy more samples during training and avoid overfitting. Our experiments showed that for a 33million bilingual SCFG rules statistical machine translation system, the autoencoder generalizes much better than sparse features alone using the same optimization framework.

We present a novel approach for improving communication success between users of speech-to-speech translation systems by automatically detecting errors in the output of automatic speech recognition (ASR) and statistical machine translation (SMT) systems. Our approach initiates system-driven targeted clarification about errorful regions in user input and repairs them given user responses. Our system has been evaluated by unbiased subjects in live mode, and results show improved success of communication between users of the system.

The paper mainly focuses on the rules of nitrogen transferring and transformation under the condition of water logging. Our experiment had set three groups, waterlogged in 1 day, waterlogged in 3 days, waterlogged in 5days. Through taking water samples in testing holes, testing the concentration of ammonium, nitrate, TN to study the rules of nitrogen transferring and transformation under the condition of water logging. Experiments had exposed that, nitrogen appeared in where roots grow vigorously in upland field. During the first day to the third day of waterlogged, the concentration of nitrogen declined steeply in the underground 10cm-80cm. After that, the level of nitrogen stabilized about some concentration. The amount of nitrogen absorbed by cotton lowered after 3 days of water logging.

To achieve the product green design, from the view point of disassembly after the product retirement, the article studies disassembly sequence of products and indexes related to disassembly process respectively and makes a comprehensive evaluation. And on the basis, it puts forward reasonable and feasible measures. Finally it combines examples to verify the methods.

Mental stress has been proved to play an important role in civilization diseases; how to improve the quality of diagnosis has become an important task. In this paper, we propose a hybrid evolutionary approach, RS-HTGA, to extract knowledge to support the physicians' decision-making. The proposed method has been successfully applied to metal stress biomedical signal diagnosis and clinical data sets. The results show that the proposed method can not only effectively extract the decision rules without external information or prior knowledge, but also allowed expert reasoning. The experimental results also show that the model can achieve a higher level of accuracy (overall accuracy of 60%, coverage of 100%).

We describe our recent effort implementing SRI's UMPC-based Pashto speech-to-speech (S2S) translation system on a smart phone running the Android operating system. In order to maintain very low latencies of system response on computationally limited smart phone platforms, we developed efficient algorithms and data structures and optimized model sizes for various system components. Our current Android-based S2S system requires less than one-fourth the system memory and significantly lower processor speed with a sacrifice of 15% relative loss of system accuracy, compared to a laptop-based platform.

In this paper, a novel all phase DFT filter bank (APFB) for multi-carrier modulation is presented and the efficient implementation structure is derived. A narrowband APDFT prototype filter exhibits higher stop attenuation and an M-band APFB with the strict power complementarities can be easily achieved by designing the frequency response vectors. The impulse response of the proposed APFB is also real and symmetric. For a modulation system with the sub-channel number M, the overlapped factor 2q and the symbol period for sub channel T 0 , the multiplication processing unit (MPU) amount is (3Mlog 2 M+2qMlog 2 2q+6qM+M+1)/T 0 per second for the transmitter and (2qM+2qMlog 2 2qM+4M−4)/T 0 per second for the receiver with the efficient structure whereas the MPU amount is (4qM3-M2)/T 0 per second for those with the direct realization structure. The comparison shows that the proposed structure possesses the advantage of much higher computational efficiency than the direct realization structure.

Entire lifecycle information integration model includes the knowledge of entire lifecycle, which can produce design model, manufacture model, quality checking model and maintenance model. In the paper, entire lifecycle information integration model is presented to realize the share of entire lifecycle information. Finally, data structure of entire lifecycle information integration model is created and entire lifecycle data of ship are included in the data structure.

The information integration model which includes all the knowledge in the manufacture phase of produce is very significant for produce manufacture. In the study, the construction of information integration model based on the design knowledge of product is presented to gain manufacture information completely and fast. Firstly, the information integration model is introduced, and the transformation of information integration model by design model is described. In the transformation process, the manufacture structure is transformed by the structure of design. Finally, the data structure of information integration model is described. By the analysis of the experiments, the method of construction of information integration model based on the design knowledge is feasible and effective.

Design and maintenance integration model is presented to provide diagnosis and maintenance knowledge of ship. The design and maintenance integration BOM model can produce design BOM view and maintenance BOM view by mapping of design and maintenance integration BOM model. Firstly, the structure of design and maintenance integration model is given, and design BOM view and maintenance BOM view are described. Secondly, design BOM view and maintenance BOM view are described. The structure of product maintenance expresses the function relationship of different layers in the structure of product and maintenance layer relationship, and the maintenance knowledge of the node in the structure of product maintenance includes replaceable component and failure knowledge. Finally, the system based on the maintenance BOM model which derives from design and maintenance integration model is realized.