A5074346065- Speech and Audio Processing
- Image and Video Quality Assessment
- Advanced Vision and Imaging
- Advanced Image Processing Techniques
- Neural Networks and Applications
- Hearing Loss and Rehabilitation
- Music and Audio Processing
- Video Coding and Compression Technologies
- Cryptographic Implementations and Security
- Chaos-based Image/Signal Encryption
- Image Enhancement Techniques
- Image and Signal Denoising Methods
- Robotics and Sensor-Based Localization
- Product Development and Customization
- Reinforcement Learning in Robotics
- Advanced Adaptive Filtering Techniques
- Optical measurement and interference techniques
- Image Processing Techniques and Applications
- Model Reduction and Neural Networks
- Multimedia Communication and Technology
- Manufacturing Process and Optimization
- Anomaly Detection Techniques and Applications
- Video Analysis and Summarization
- Coding theory and cryptography
- Advanced Data Compression Techniques
Technical University of Munich
2016-2025
Automatic Data Processing (United States)
2022
Lohmann (Germany)
2013-2016
Ruhr University Bochum
2010
Information Technology Institute
2009
Information Technology University
2009
Munich University of Applied Sciences
2008
Ludwig-Maximilians-Universität München
2006-2007
Abstract The advances in reinforcement learning have recorded sublime success various domains. Although the multi-agent domain has been overshadowed by its single-agent counterpart during this progress, gains rapid traction, and latest accomplishments address problems with real-world complexity. This article provides an overview of current developments field deep learning. We focus primarily on literature from recent years that combines methods a scenario. To survey works constitute...
To address control difficulties in laser welding, we propose the idea of a self-learning and self-improving welding system that combines three modern machine learning techniques. We first show ability deep neural network to extract meaningful, low-dimensional features from high-dimensional laser-welding camera data. These are then used by temporal-difference algorithm predict anticipate important aspects system's sensor The third part our proposed architecture suggests using these...
Quality of Experience (QoE) in multimedia applications is closely linked to the end users' perception and therefore its assessment requires subjective user studies order evaluate degree delight or annoyance as experienced by users. QoE crowdtesting refers using crowdsourcing, where anonymous test subjects conduct tests remotely their preferred environment. The advantages lie not only reduced time costs for tests, but also a large diverse panel international, geographically distributed users...
Exploiting a priori known structural information lies at the core of many image reconstruction methods that can be stated as inverse problems. The synthesis model, which assumes images decomposed into linear combination very few atoms some dictionary, is now well established tool for design algorithms. An interesting alternative analysis where signal multiplied by an operator and outcome assumed to sparse. This approach has only recently gained increasing interest. quality based on model...
Machine intelligence already helps medical staff with a number of tasks. Ethical decision-making, however, has not been handed over to computers. In this proof-of-concept study, we show how an algorithm based on Beauchamp and Childress' prima-facie principles could be employed advise range moral dilemma situations that occur in institutions. We explain why chose fuzzy cognitive maps set up the advisory system utilized machine learning train it. report difficult task operationalizing...
Today's manufacturing and assembly systems have to be flexible adapt quickly an increasing number variety of products, changing market volumes. To manage these dynamics, several production concepts (e.g., flexible, reconfigurable, changeable or autonomous systems) were proposed partly realized in the past years. This paper presents general principles autonomy concepts, methods technologies realize cognitive planning, control operation systems. Starting with introduction on historical context...
In this work we propose a method for estimating disparity maps from very few measurements. Based on the theory of Compressive Sensing, our algorithm accurately reconstructs only using about 5% entire map. We conjugate subgradient arising optimization problem that is applicable to large scale systems and recovers map efficiently. Experiments are provided show effectiveness proposed approach robust behavior under noisy conditions.
Video quality assessment with subjective testing is both time consuming and expensive. An interesting new approach to traditional the so-called crowdsourcing, moving effort into internet. We therefore propose in this contribution QualityCrowd framework effortlessly perform crowdsourcing. allows codec independent a simple web interface, usable common browsers. compared results from an online test using standardized environment. This comparison shows that delivers equivalent within acceptable...
Manufacturing processes have undergone tremendous technological progress in recent decades. To meet the agile philosophy industry, data-driven algorithms need to handle growing complexity, particularly Computer Numerical Control machining. enhance scalability of machine learning real-world applications, this paper presents a benchmark dataset for process monitoring brownfield milling machines based on acceleration data. The data is collected from production plant using smart collection...
One of the major reasons why elderly lose their ability to live independently at home is decline in gait performance. A measure assess performance using accelerometers step counting. The main problem with most detection algorithms loss accuracy low speeds ( 0.8 m/s) which limits use frail populations. In this paper, a algorithm was developed and validated data from 10 healthy adults 21 institutionalized seniors, predominantly older adults. Data were recorded single waist-worn triaxial...
The clinical spectrum of acute SARS-CoV-2 infection ranges from an asymptomatic to life-threatening disease. Considering the broad severity, reliable biomarkers are required for early risk stratification and prediction outcomes. Despite numerous efforts, no COVID-19-specific biomarker has been established guide further diagnostic or even therapeutic approaches, most likely due insufficient validation, methodical complexity, economic factors. COVID-19-associated coagulopathy is a hallmark...
The research on video quality metrics depends the results from subjective testing for both design and development of metrics, but also their verification. As it is often too cumbersome to conduct tests, freely available data sets that include mean opinion scores distorted videos are becoming ever more important. While many datasets already widely available, majority these focus smaller resolutions. We therefore present in this contribution TUM high definition 1080p25 1080p50, encoded with...
A modern technique for robotic sound source detection using a dataset Head-Related Transfer Functions (HRTFs) is presented. To ensure fast detection, the HRTFs are reduced three different techniques, namely; Diffuse-Field Equalization, Balanced Model Truncation, and Principle Component Analysis. new criterion introduced to be satisfied by set of output signals from microphones dummy robot head. This then used find location in accordance with HRTF datasets. The suggested method verified...
No-reference video quality metrics are becoming ever more popular, as they useful in real-life applications compared to full-reference metrics. Many proposed extract features related human perception from the individual frames. Hence sequences have be decoded first, before can applied. In order avoid decoding just for estimation, we therefore present this contribution a no-reference metric HDTV that uses directly extracted H.264/AVC bitstream. We combine these with results subjective tests...
In the last decades, data-driven methods have gained great popularity in industry, supported by state-of-the-art advancements machine learning. These require a large quantity of labeled data, which is difficult to obtain and mostly costly challenging. To address these challenges, researchers turned their attention unsupervised few-shot learning methods, produced encouraging results, particularly areas computer vision natural language processing. With lack pretrained models, time series...
This contribution presents a no-reference video quality metric, which is based on set of simple rules that assigns given to one four different content classes. The classes distinguish between sequences are coded with very low data rate, sensitive blocking effects, blurring, and general model for all other types sequences. appropriate class sequence selected the evaluation feature values an additional low-quality version video, generated by encoding. visual estimated using features, includes...
The interest in content adaptive mesh generation of images has been arising lately due to its wide area applications image processing. major issue is represent an with a low number pixels while preserving content. These or the nonuniform samples are then used generate that approximates corresponding image. This work presents novel method based on Binary Space Partitions combination three clustering schemes approximate mesh. algorithm ability simultaneously reduce and approximation. idea...
Video quality evaluation with subjective testing is both time consuming and expensive. A promising new approach to traditional the so-called crowdsourcing, moving effort into Internet. The advantages of this are not only access a larger more diverse pool test subjects, but also significant reduction financial burden. Recent contributions have shown that crowd-based video assessment can deliver results comparable in some cases. In general, however, problems arise, as no longer every detail be...
For sound localization methods to be useful in real-time scenarios, the processing power requirements must low enough allow real time of audio inputs. We propose a new binaural source technique based on using only two microphones placed inside ear canal robot dummy head. The head is equipped with artificial ears and mounted torso. In contrast existing 3D microphone arrays, our novel method employs simple correlation approach generic set related transfer functions (HRTFs). proposed...
We propose in this paper a new scheme based on compressed sensing to compress depth map. first subsample the entity frequency domain take advantage of its compressibility. then derive reconstruction recover original map from subsamples using non-linear conjugate gradient minimization scheme. preserve discontinuities at edges and ensure smoothness elsewhere by incorporating Total Variation constraint minimization. The results we obtained various test maps show that proposed method leads lower...