Refine
Document Type
- Book chapter (12)
- Conference proceeding (7)
- Journal article (4)
- Book (2)
- Anthology (1)
Is part of the Bibliography
- yes (26)
Institute
- Technik (26) (remove)
Publisher
- Springer (26) (remove)
With the rapid development of globalization, the demand for translation between different languages is also increasing. Although pre-training has achieved excellent results in neural machine translation, the existing neural machine translation has almost no high-quality suitable for specific fields. Alignment information, so this paper proposes a pre-training neural machine translation with alignment information via optimal transport. First, this paper narrows the representation gap between different languages by using OTAP to generate domain-specific data for information alignment, and learns richer semantic information. Secondly, this paper proposes a lightweight model DR-Reformer, which uses Reformer as the backbone network, adds Dropout layers and Reduction layers, reduces model parameters without losing accuracy, and improves computational efficiency. Experiments on the Chinese and English datasets of AI Challenger 2018 and WMT-17 show that the proposed algorithm has better performance than existing algorithms.
In clothing e-commerce, the challenge of optimally recommending clothing that suits a user’s unique characteristics remains a pressing issue. Many platforms simply recommend best-selling or popular clothing, without taking into account important attributes like user’s face color, pupil color, face shape, age, etc. To solve this problem, this paper proposes a personalized clothing recommendation algorithm that incorporates the established 4-Season Color System and user-specific biological characteristics. Firstly, the attributes and colors of clothing are classified by Fnet network, that can learn disjoint label combinations and mitigate the issue of excessive labels. Secondly, on the basis of the 4-Season Color System, the user’s face color model is trained by combined MobileNetV3_DTL, which ensures the model’s generalization and improves the training speed. Thirdly, user’s face shape and age are divided into different categories by an Inception network. Finally, according to the users’ face color, age, face shape and other information, personalized clothing is recommended in a coarse-to-fine manner. Experiments on five datasets demonstrate that the algorithm proposed in this paper achieves state-of-the-art results.
Silicon neurons represent different levels of biological details and accuracies as a trade-off between complexity and power consumption. With respect to this trade-off and high similarity to neuron behaviour models, relaxation-type oscillator circuits often yield a good compromise to emulate neurons. In this chapter, two exemplified relaxation-type silicon neurons are presented that emulate neural behaviour with energy consumption under the scale of nJ/spike. The first proposed fully CMOS relaxation SiN is based on mathematical Izhikevich model and can mimic a broad range of physiologically observable spike patterns. The results of kinds of biologically plausible output patterns and coupling process of two SiNs are presented in 0.35 μm CMOS technology. The second type is a novel ultra-low-frequency hybrid CMOS-memristive SiN based on relaxation oscillators and analog memristive devices. The hybrid SiN directly emulates neuron behaviour in the range of physiological spiking frequencies (less than 100 Hz). The relaxation oscillator is implemented and fabricated in 0.13 μm CMOS technology. An autonomous neuronal synchronization process is demonstrated with two relaxation oscillators coupled by an analog memristive device in the measurement to emulate the synchronous behaviour between spiking neurons.
In recent years, 3D facial reconstructions from single images have garnered significant interest. Most of the approaches are based on 3D Morphable Model (3DMM) fitting to reconstruct the 3D face shape. Concurrently, the adoption of Generative Adversarial Networks (GAN) has been gaining momentum to improve the texture of reconstructed faces. In this paper, we propose a fundamentally different approach to reconstructing the 3D head shape from a single image by harnessing the power of GAN. Our method predicts three maps of normal vectors of the head’s frontal, left, and right poses. We are thus presenting a model-free method that does not require any prior knowledge of the object’s geometry to be reconstructed.
The key advantage of our proposed approach is the substantial improvement in reconstruction quality compared to existing methods, particularly in the case of facial regions that are self-occluded in the input image. Our method is not limited to 3d face reconstruction. It is generic and applicable to multiple kinds of 3D objects. To illustrate the versatility of our method, we demonstrate its efficacy in reconstructing the entire human body.
By delivering a model-free method capable of generating high-quality 3D reconstructions, this paper not only advances the field of 3D facial reconstruction but also provides a foundation for future research and applications spanning multiple object types. The implications of this work have the potential to extend far beyond facial reconstruction, paving the way for innovative solutions and discoveries in various domains.
The aim of this work is the development of artificial intelligence (AI) application to support the recruiting process that elevates the domain of human resource management by advancing its capabilities and effectiveness. This affects recruiting processes and includes solutions for active sourcing, i.e. active recruitment, pre-sorting, evaluating structured video interviews and discovering internal training potential. This work highlights four novel approaches to ethical machine learning. The first is precise machine learning for ethically relevant properties in image recognition, which focuses on accurately detecting and analysing these properties. The second is the detection of bias in training data, allowing for the identification and removal of distortions that could skew results. The third is minimising bias, which involves actively working to reduce bias in machine learning models. Finally, an unsupervised architecture is introduced that can learn fair results even without ground truth data. Together, these approaches represent important steps forward in creating ethical and unbiased machine learning systems.
Dieses Buch vermittelt die grundlegenden Kenntnisse für den Layoutentwurf digitaler und analoger Schaltungen. Neben den ingenieurwissenschaftlichen Grundlagen werden auch Werkzeugaspekte behandelt. Das Werk befähigt Ingenieure, einen Schaltplan oder eine Netzliste in eine Layoutdarstellung zur Fertigung eines integrierten Schaltkreises (IC) oder einer Leiterplatte umzusetzen. Nach einer einleitenden Übersicht zu Fertigungstechnologien, Besonderheiten der Mikroelektronik und den Aufgaben des Layoutentwurfs behandelt Kap. 2 zunächst die technologischen Grundlagen der IC-Fertigung. Darauf aufbauend werden nachfolgend alle Aspekte des Layoutentwurfs vertieft: Schnittstellen, Entwurfsregeln und Bibliotheken (Kap. 3), Entwurfsstile, -modelle und -flüsse (Kap. 4), Entwurfsschritte (Kap. 5), Besonderheiten des analogen IC-Entwurfs (Kap. 6) und schließlich Zuverlässigkeitsmaßnahmen (Kap. 7). Das Buch eignet sich als Lehrbuch in den Ingenieurwissenschaften und als Nachschlagewerk für Schaltungs- und Layoutentwickler in der Industrie.
Purpose
Injury or inflammation of the middle ear often results in the persistent tympanic membrane (TM) perforations, leading to conductive hearing loss (HL). However, in some cases the magnitude of HL exceeds that attributable by the TM perforation alone. The aim of the study is to better understand the effects of location and size of TM perforations on the sound transmission properties of the middle ear.
Methods
The middle ear transfer functions (METF) of six human temporal bones (TB) were compared before and after perforating the TM at different locations (anterior or posterior lower quadrant) and to different degrees (1 mm, ¼ of the TM, ½ of the TM, and full ablation). The sound-induced velocity of the stapes footplate was measured using single-point laser-Doppler-vibrometry (LDV). The METF were correlated with a Finite Element (FE) model of the middle ear, in which similar alterations were simulated.
Results
The measured and calculated METF showed frequency and perforation size dependent losses at all perforation locations. Starting at low frequencies, the loss expanded to higher frequencies with increased perforation size. In direct comparison, posterior TM perforations affected the transmission properties to a larger degree than anterior perforations. The asymmetry of the TM causes the malleus-incus complex to rotate and results in larger deflections in the posterior TM quadrants than in the anterior TM quadrants. Simulations in the FE model with a sealed cavity show that small perforations lead to a decrease in TM rigidity and thus to an increase in oscillation amplitude of the TM mainly above 1 kHz.
Conclusion
Size and location of TM perforations have a characteristic influence on the METF. The correlation of the experimental LDV measurements with an FE model contributes to a better understanding of the pathologic mechanisms of middle-ear diseases. If small perforations with significant HL are observed in daily clinical practice, additional middle ear pathologies should be considered. Further investigations on the loss of TM pretension due to perforations may be informative.
We discuss the fabrication technologies for IC chips in this chapter. We will focus on the main process steps and especially on those aspects that are of particular importance for understanding how they affect, and in some cases drive, the layout of ICs. All our analyses in this chapter will be for silicon as the base material; the principles and understanding gained can be applied to other substrates as well. Following a brief introduction to the fundamentals of IC fabrication (Sect. 2.1) and the base material used in it, namely silicon (Sect. 2.2), we discuss the photolithography process deployed for all structuring work in Sect. 2.3. We will then present in Sect. 2.4 some theoretical opening remarks on typical phenomena encountered in IC fabrication. Knowledge of these phenomena is very useful for understanding the process steps we cover in Sects. 2.5–2.8. We examine a simple exemplar process in Sect. 2.9 and observe how a field-effect transistor (FET) – the most important device in modern integrated circuits—is created. To drive the key points home, we provide a review of each topic at the end of every section from the point of view of layout design by discussing relevant physical design aspects.
Investigation of tympanic membrane influences on middle-ear impedance measurements and simulations
(2020)
This study simulates acoustic impedance measurements in the human ear canal and investigates error influences due to improperly accounted evanescence in the probe’s near field, cross-section area changes, curvature of the ear canal, and pressure inhomogeneities across the tympanic membrane, which arise mainly at frequencies above 10 kHz. Evanescence results from strongly damped modes of higher order, which can only be found in the near field of the sound source and are excited due to sharp cross-sectional changes as they occur at the transition from the probe loudspeaker to the ear canal. This means that different impedances are measured depending on the probe design. The influence of evanescence cannot be eliminated completely from measurements, however, it can be reduced by a probe design with larger distance between speaker and microphone. A completely different approach to account for the influence of evanescence is to evaluate impedance measurements with the help of a finite element model, which takes the precise arrangement of microphone and speaker in the measurement into account. The latter is shown in this study exemplary on impedance measurements at a tube terminated with a steel plate. Furthermore, the influences of shape changes of the tympanic membrane and ear canal curvature on impedance are investigated.
This study describes a non-contact measuring and system identification procedure for evaluating inhomogeneous stiffness and damping characteristics of the annular ligament in the physiological amplitude and frequency range without the application of large static external forces that can cause unnatural displacements of the stapes. To verify the procedure, measurements were first conducted on a steel beam. Then, measurements on an individual human cadaveric temporal bone sample were performed. The estimated results support the inhomogeneous stiffness and damping distribution of the annular ligament and are in a good agreement with the multiphoton microscopy results which show that the posterior-inferior corner of the stapes footplate is the stiffest region of the annular ligament.
This book covers the fundamental knowledge of layout design from the ground up, addressing both physical design, as generally applied to digital circuits, and analog layout. Such knowledge provides the critical awareness and insights a layout designer must possess to convert a structural description produced during circuit design into the physical layout used for IC/PCB fabrication.
Efficient and robust 3D object reconstruction based on monocular SLAM and CNN semantic segmentation
(2019)
Various applications implement slam technology, especially in the field of robot navigation. We show the advantage of slam technology for independent 3d object reconstruction. To receive a point cloud of every object of interest void of its environment, we leverage deep learning. We utilize recent cnn deep learning research for accurate semantic segmentation of objects. In this work, we propose two fusion methods for cnn-based semantic segmentation and slam for the 3d reconstruction of objects of interest in order to obtain a more robustness and efficiency. As a major novelty, we introduce a cnn-based masking to focus slam only on feature points belonging to every single object. Noisy, complex or even non-rigid features in the background are filtered out, improving the estimation of the camera pose and the 3d point cloud of each object. Our experiments are constrained to the reconstruction of industrial objects. We present an analysis of the accuracy and performance of each method and compare the two methods describing their pros and cons.
This paper presents a novel multi-modal CNN architecture that exploits complementary input cues in addition to sole color information. The joint model implements a mid-level fusion that allows the network to exploit cross modal interdependencies already on a medium feature-level. The benefit of the presented architecture is shown for the RGB-D image understanding task. So far, state-of-the-art RGB-D CNNs have used network weights trained on color data. In contrast, a superior initialization scheme is proposed to pre-train the depth branch of the multi-modal CNN independently. In an end-to-end training the network parameters are optimized jointly using the challenging Cityscapes dataset. In thorough experiments, the effectiveness of the proposed model is shown. Both, the RGB GoogLeNet and further RGB-D baselines are outperformed with a significant margin on two different tasks: semantic segmentation and object detection. For the latter, this paper shows how to extract object level groundtruth from the instance level annotations in Cityscapes in order to train a powerful object detector.
A 3D face modelling approach for pose-invariant face recognition in a human-robot environment
(2017)
Face analysis techniques have become a crucial component of human-machine interaction in the fields of assistive and humanoid robotics. However, the variations in head-pose that arise naturally in these environments are still a great challenge. In this paper, we present a real-time capable 3D face modelling framework for 2D in-the-wild images that is applicable for robotics. The fitting of the 3D Morphable Model is based exclusively on automatically detected landmarks. After fitting, the face can be corrected in pose and transformed back to a frontal 2D representation that is more suitable for face recognition. We conduct face recognition experiments with non-frontal images from the MUCT database and uncontrolled, in the wild images from the PaSC database, the most challenging face recognition database to date, showing an improved performance. Finally, we present our SCITOS G5 robot system, which incorporates our framework as a means of image pre-processing for face analysis.
Broad acceptance of finite-element-based analysis of structural problems and the increased availability of CAD-systems for structural tasks, which help to generate meshes of non-trivial geometries, have been setting a standard for the evaluation of designs in mechanical engineering in the last few decades. The development of automated or semi-automated optimizers, integrated into the Computer-Aided Engineering (CAE)-packages or working as outer loop machines, requiring the solver to do the analysis of the specific designs, has been accepted by most advanced users of the simulation community as well. The availability and inexpensive processing power of computers is increasing without any limitations foreseen in the coming years. There is little doubt that virtual product development will continue using the tools that have proved to be so successful and so easy to handle.
Current fields of interest
(2016)
If we review the research done in the field of optimization, the following topics appear to be the focus of current development:
– Optimization under uncertainties, taking into account the inevitable scatter of parts, external effects and internal properties. Reliability and robustness both have to be taken into account when running optimizations, so the name Robust Design Optimization (RDO) came into use.
– Multi-Objective Optimization (MOO) handles situations in which different participants in the development process are developing in different directions. Typically we think of commercial and engineering aspects, but other constellations have to be looked at as well, such as comfort and performance or price and consumption.
– Process development of the entire design process, including optimization from early stages, might help avoid inefficient efforts. Here the management of virtual development has to be re-designed to fit into a coherent scheme.
...
There are many other fields where interesting progress is being made. We limit our discussion to the first three questions.
To illustrate the power and the pitfalls of Bionic Optimization, we will show some examples spanning classes of applications and employing various strategies. These applications cover a broad range of engineering tasks. Nevertheless, there is no guarantee that our experiences and our examples will be sufficient to deal with all questions and issues in a comprehensive way. As general rule it might be stated, that for each class of problems, novices should begin with a learning phase. So, in this introductory phase, we use simple and quick examples, e.g., using small FE-models, linear load cases, short time intervals and simple material models. Here beginners within the Bionic Optimization community can learn which parameter combinations to use. In Sect. 3.3 we discuss strategies for optimization study acceleration. Making use of these parameters as starting points is one way to set the specific ranges, e.g., number of parents and kids, crossing, mutation radii and, numbers of generations. On the other hand, these trial runs will doubtless indicate that Bionic Optimization needs large numbers of individual designs, and considerable time and computing power. We recommend investing enough time preparing each task in order to avoid the frustration should large jobs fail after long calculation times.
Application to CAE systems
(2016)
Due to the broad acceptance of CAD-systems based on 3D solids, the geometric data of all common CAE (Computer-Aided Engineering) software, at least in mechanical engineering, are based on these solids. We use solid models, where the space filled by material is defined in a simple and easily useable way. Solid models allow for the development of automated meshers that transform solid volumes into finite elements. Even after some unacceptable initial trials, users are able to generate meshes of non-trivial geometries within minutes to hours, instead of days or weeks. Once meshing had no longer been the cost limiting factor of finite element studies, numerical simulation became a tool for smaller industries as well.
Due to the broad acceptance of CAD-systems based on 3D solids , the geometric data of all common CAE (Computer-Aided Engineering) software, at least in mechanical engineering, are based on these solids. We use solid models , where the space filled by material is defined in a simple and easily useable way. Solid models allow for the development of automated meshers that transform solid volumes into finite elements. Even after some unacceptable initial trials, users are able to generate meshes of non-trivial geometries within minutes to hours, instead of days or weeks. Once meshing had no longer been the cost limiting factor of finite element studies, numerical simulation became a tool for smaller industries as well.
In the early days of automated meshing development, there were discussions over the use of tetragonal (Fig. 4.1) or hexagonal based meshes. But, after a short period of time, it became evident, that there were and will always be many problems using automated meshers to generate hexagonal elements . So today nearly all automated 3D-meshing systems use tetragonal elements .
We have seen that bionic optimization can be a powerful tool when applied to problems with non-trivial landscapes of goals and restrictions. This, in turn, led us to a discussion of useful methodologies for applying this optimization to real problems. On the other hand, it must be stated that each optimization is a time consuming process as soon as the problem expands beyond a small number of free parameters related to simple parabolic responses. Bionic optimization is not a quick approach to solving complex questions within short times. In some cases it has the potential to fail entirely, either by sticking to local maxima or by random exploration of the parameter space without finding any promising solutions. The following sections present some remarks on the efficiency and limitations users must be aware of. They aim to increase the knowledge base of using and encountering bionic optimization. But they should not discourage potential users from this promising field of powerful strategies to find good or even the best possible designs.
In this chapter we introduce methods to improve mechanical designs by bionic methods. In most cases we assume that a general idea of the part or system is given by a set of data or parameters. Our task is to modify these free parameters so that a given goal or objective is optimized without violation of any of the existing restrictions.
Motivation
(2016)
Since human beings started to work consciously with their environment, they have tried to improve the world they were living in. Early use of tools, increasing quality of these tools, use of new materials, fabrication of clay pots, and heat treatment of metals: all these were early steps of optimization. But even on lower levels of life than human beings or human society, we find optimization processes. The organization of a herd of buffalos to face their enemies, the coordinated strategies of these enemies to isolate some of the herd’s members, and the organization of bird swarms on their long flights to their winter quarters: all these social interactions are optimized strategies of long learning processes, most of them the result of a kind of collective intelligence acquired during long selection periods.
The book provides suggestions on how to start using bionic optimization methods, including pseudo-code examples of each of the important approaches and outlines of how to improve them. The most efficient methods for accelerating the studies are discussed. These include the selection of size and generations of a study’s parameters, modification of these driving parameters, switching to gradient methods when approaching local maxima, and the use of parallel working hardware.
Bionic optimization means finding the best solution to a problem using methods found in nature. As evolutionary strategies and particle swarm optimization seem to be the most important methods for structural optimization, we primarily focus on them. Other methods such as neural nets or ant colonies are more suited to control or process studies, so their basic ideas are outlined in order to motivate readers to start using them.
A set of sample applications shows how bionic optimization works in practice. From academic studies on simple frames made of rods to earthquake-resistant buildings, readers follow the lessons learned, difficulties encountered and effective strategies for overcoming them. For the problem of tuned mass dampers, which play an important role in dynamic control, changing the goal and restrictions paves the way for multi-objective-optimization. As most structural designers today use commercial software such as FE-Codes or CAE systems with integrated simulation modules, ways of integrating bionic optimization into these software packages are outlined and examples of typical systems and typical optimization approaches are presented.
The closing section focuses on an overview and outlook on reliable and robust as well as on multi-objective optimization, including discussions of current and upcoming research topics in the field concerning a unified theory for handling stochastic design processes.
This article covers the design of highly integrated gate drivers and level shifters for high-speed, high power efficiency and dv/dt robustness with focus on automotive applications. With the introduction of the 48 V board net in addition to the conventional 12 V battery, there is an increasing need for fast switching integrated gate drivers in the voltage range of 50 V and above. State-of-the-art drivers are able to switch 50 V in less than 5 ns. The high-voltage electrical drive train demands for galvanic isolated and highly integrated gate drivers. A gate driver with bidirectional signal transmission with a 1 MBit/s amplitude modulation, 10/20 MHz frequency modulation and power transfer over one single transformer will be discussed. The concept of high-voltage charge storing enables an area-efficient fully integrated bootstrapping supply with 70 % less area consumption. EMC is a major concern in automotive. Gate drivers with slope control optimize EMC while maintaining good switching efficiency. A current mode gate driver, which can change its drive current within 10 ns, results in 20 dBuV lower emissions between 7 and 60 MHz and 52 % lower switching loss compared to a conventional constant current gate driver.
Virtual prototyping of integrated mixed-signal smart sensor systems requires high-performance co-simulation of analog frontend circuitry with complex digital controller hardware and embedded real-time software. We use SystemC/TLM 2.0 in conjunction with a cycle-count accurate temporal decoupling approach (TD) to simulate digital components and firmware code execution at high speed while preserving clock-cycle accuracy and, thus, real-time behavior at time quantum boundaries. Optimal time quanta ensuring real-time capability can be calculated and set automatically during simulation if the simulation engine has access to exact timing information about upcoming inter-process communication events. These methods fail in the case of non-deterministic, asynchronous events, resulting in potentially invalid simulation results. In this paper, we propose an extension to the case of asynchronous events generated by blackbox sources from which a priori event timing information is not available, such as coupled analog simulators or hardware in the loop. Additional event processing latency or rollback effort caused by temporal decoupling is minimized by calculating optimal time quanta dynamically in a SystemC model using a linear prediction scheme. We analyze the theoretical performance of the presented predictive temporal decoupling approach (PTD) by deriving a cost model that expresses the expected simulation effort in terms of key parameters such as time quantum size and CPU time per simulation cycle. For an exemplary smart-sensor system model, we show that quasi-periodic events that trigger activities in TD processes are handled accurately after the predictor has settled.
Despite 30 years of Electronic Design Automation, analog IC layouts are still handcrafted in a laborious fashion today due to the complex challenge of considering all relevant design constraints. This paper presents Self-organized Wiring and Arrangement of Responsive Modules (SWARM), a novel approach addressing the problem with a multi-agent system: autonomous layout modules interact with each other to evoke the emergence of overall compact arrangements that fit within a given layout zone. SWARM´s unique advantage over conventional optimization-based and procedural approaches is its ability to consider crucial design constraints both explicitly and implicitly. Several given examples show that by inducing a synergistic flow of self-organization, remarkable layout results can emerge from SWARM’s decentralized decision-making model.
The capability of the method of Immersion transmission ellipsometry (ITE) (Jung et al. Int Patent WO, 2004/109260) to not only determine three-dimensional refractive indices in anisotropic thin films (which was already possible in the past), but even their gradients along the z-direction (perpendicular to the film plane) is investigated in this paper. It is shown that the determination of orientation gradients in deep-sub-lm films becomes possible by applying ITE in combination with reflection ellipsometry. The technique is supplemented by atomic force microscopy for measuring the film thickness. For a photooriented thin film, no gradient was found, as expected. For a photo-oriented film, which was subsequently annealed in a nematic liquid crystalline phase, an order was found similar to the one applied in vertically aligned nematic displays, with a tilt angle varying along the z-direction. For fresh films, gradients were only detected for the refractive index perpendicular to the film plane, as expected.