Marc Kurz
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Marc Kurz has a total of 25 co-authors. Here are the top co-authors by number of shared publications:
- Erik Sonnleitner
Department for Smart and Interconnected Living, University of Applied Sciences Upper Austria, Softwarepark 11, 4232 Hagenberg, Austria
10 shared publications - Alexander Palmanshofer
Department of Mobility and Energy, University of Applied Sciences Upper Austria, 4232 Hagenberg, Austria
2 shared publications - Jens Krösche
University of Applied Sciences Upper Austria
2 shared publications - Stephan Selinger
Department for Smart and Interconnected Living, University of Applied Sciences Upper Austria, Softwarepark 11, 4232 Hagenberg, Austria
2 shared publications - Andreas Roither
Department of Mobility and Energy, University of Applied Sciences Upper Austria, 4232 Hagenberg, Austria
1 shared publication - Aya Mohamed
Institute for Application-oriented Knowledge Processing (FAW)
1 shared publication
Latest Publications
- 9 October 2026
- Published by Springer Nature
- p. 412-430
An AI-Driven Transparency Database for Cross-National Comparison of European Electricity Tariffs
- 9 October 2026
- Published by Springer Nature
- p. 412-430
An AI-Driven Transparency Database for Cross-National Comparison of European Electricity Tariffs
- 8 February 2026
- Published by Springer Nature
- p. 294-311
Towards User-Centric Authorization for Data Access, Sharing and Control
- 8 February 2026
- Published by Springer Nature
- p. 294-311
Towards User-Centric Authorization for Data Access, Sharing and Control
- Open Access
- 4 January 2026
- Published by MDPI AG in Information
- Vol. 17 (1) , 40
Listen Closely: Self-Supervised Phoneme Tracking for Children’s Reading Assessment
Reading proficiency in early childhood is crucial for academic success and intellectual development. However, more and more children are struggling with reading. According to the last PISA study in Austria, one out of five children is dealing with reading difficulties. The reasons for this are diverse, but an application that tracks children while reading aloud and guides them when they experience difficulties could offer meaningful help. Therefore, this proposal explores a prototyping approach for a core component that tracks children’s reading using a self-supervised Wav2Vec2 model with a limited amount of data. Self-supervised learning allows models to learn general representations from large amounts of unlabeled audio, which can then be fine-tuned on smaller, task-specific datasets, making it especially useful when labeled data is limited. Our model is operating on the phonetic level with the help of the International Phonetic Alphabet (IPA). To implement this, the KidsTALC dataset from the Leibniz University Hannover was used, which contains spontaneous speech recordings of German-speaking children. To enhance the training data and improve robustness, several data augmentation techniques were applied and evaluated, including pitch shifting, formant shifting, and speed variation. The models were trained using different data configurations to compare the effects of data variety and quality on recognition performance. The best model trained in this work achieved a phoneme error rate (PER) of 14.3% and a word error rate (WER) of 31.6% on unseen child speech data, demonstrating the potential of self-supervised models for such use cases.
- 4 January 2026
- Published by MDPI AG in Information
- Vol. 17 (1) , 40
Open AccessListen Closely: Self-Supervised Phoneme Tracking for Children’s Reading Assessment
Reading proficiency in early childhood is crucial for academic success and intellectual development. However, more and more children are struggling with reading. According to the last PISA study in Austria, one out of five children is dealing with reading difficulties. The reasons for this are diverse, but an application that tracks children while reading aloud and guides them when they experience difficulties could offer meaningful help. Therefore, this proposal explores a prototyping approach for a core component that tracks children’s reading using a self-supervised Wav2Vec2 model with a limited amount of data. Self-supervised learning allows models to learn general representations from large amounts of unlabeled audio, which can then be fine-tuned on smaller, task-specific datasets, making it especially useful when labeled data is limited. Our model is operating on the phonetic level with the help of the International Phonetic Alphabet (IPA). To implement this, the KidsTALC dataset from the Leibniz University Hannover was used, which contains spontaneous speech recordings of German-speaking children. To enhance the training data and improve robustness, several data augmentation techniques were applied and evaluated, including pitch shifting, formant shifting, and speed variation. The models were trained using different data configurations to compare the effects of data variety and quality on recognition performance. The best model trained in this work achieved a phoneme error rate (PER) of 14.3% and a word error rate (WER) of 31.6% on unseen child speech data, demonstrating the potential of self-supervised models for such use cases.
- Open Access
- 21 November 2024
- Published by MDPI AG in Electronics
- Vol. 13 (23) , 4596
Know Your Grip: Real-Time Holding Posture Recognition for Smartphones
This paper introduces a model that predicts four common smartphone-holding postures, aiming to enhance user interface adaptability. It is unique in being completely independent of platform and hardware, utilizing the inertial measurement unit (IMU) for real-time posture detection based on sensor data collected around tap gestures. The model identifies whether the user is holding and operating the smartphone with one hand or using both hands in different configurations. For model training and validation, sensor time series data undergo extensive feature extraction, including statistical, frequency, magnitude, and wavelet analyses. These features are incorporated into 74 distinct sets, tested across various machine learning frameworks—k-nearest neighbors (KNN), support vector machine (SVM), and random forest (RF)—and evaluated for their effectiveness using metrics such as cross-validation scores, test accuracy, Kappa statistics, confusion matrices, and ROC curves. The optimized model demonstrates a high degree of accuracy, successfully predicting the holding hand with a 95.7% success rate. This approach highlights the potential of leveraging sensor data to improve mobile user experiences by adapting interfaces to natural user interactions.
- 21 November 2024
- Published by MDPI AG in Electronics
- Vol. 13 (23) , 4596
Open AccessKnow Your Grip: Real-Time Holding Posture Recognition for Smartphones
This paper introduces a model that predicts four common smartphone-holding postures, aiming to enhance user interface adaptability. It is unique in being completely independent of platform and hardware, utilizing the inertial measurement unit (IMU) for real-time posture detection based on sensor data collected around tap gestures. The model identifies whether the user is holding and operating the smartphone with one hand or using both hands in different configurations. For model training and validation, sensor time series data undergo extensive feature extraction, including statistical, frequency, magnitude, and wavelet analyses. These features are incorporated into 74 distinct sets, tested across various machine learning frameworks—k-nearest neighbors (KNN), support vector machine (SVM), and random forest (RF)—and evaluated for their effectiveness using metrics such as cross-validation scores, test accuracy, Kappa statistics, confusion matrices, and ROC curves. The optimized model demonstrates a high degree of accuracy, successfully predicting the holding hand with a 95.7% success rate. This approach highlights the potential of leveraging sensor data to improve mobile user experiences by adapting interfaces to natural user interactions.
- Open Access
- 27 December 2023
- Published by MDPI AG in Electronics
- Vol. 13 (1) , 117
Exemplary Ethereum Development Strategies Regarding Security and Gas-Saving
Ethereum is a rapidly evolving blockchain with new features as well as new vulnerabilities being introduced regularly. Interaction with the network is costly compared to other blockchains or traditional software systems. When starting to develop on Ethereum, a supported smart contract programming language needs to be learned, most notably Solidity. Having various pitfalls raises the question of what the best practices for the safe and efficient usage of Ethereum are. This study primarily aims to combine knowledge from existing research resources, while also introducing new approaches learned from practical smart contract development analysis and inquiry, which are subsequently compiled into lists of best practices. The most important findings are that code quality and security should be prioritized. Moreover, some simple gas-saving strategies can help to decrease interaction costs with little effort.
- 27 December 2023
- Published by MDPI AG in Electronics
- Vol. 13 (1) , 117
Open AccessExemplary Ethereum Development Strategies Regarding Security and Gas-Saving
Ethereum is a rapidly evolving blockchain with new features as well as new vulnerabilities being introduced regularly. Interaction with the network is costly compared to other blockchains or traditional software systems. When starting to develop on Ethereum, a supported smart contract programming language needs to be learned, most notably Solidity. Having various pitfalls raises the question of what the best practices for the safe and efficient usage of Ethereum are. This study primarily aims to combine knowledge from existing research resources, while also introducing new approaches learned from practical smart contract development analysis and inquiry, which are subsequently compiled into lists of best practices. The most important findings are that code quality and security should be prioritized. Moreover, some simple gas-saving strategies can help to decrease interaction costs with little effort.
- Open Access
- 18 August 2023
- Published by MDPI AG in Applied Sciences
- Vol. 13 (16) , 9391
Let Me Help You: Improving the Novice Experience of High-Performance Keyboard Layouts with Visual Clues
Since the advent of smartphones, work tasks shift progressively towards mobile phones. The main hurdle of this progression is the substantially slower input speed of smartphones in comparison with physical keyboards. This input speed could be improved by using keyboard layouts optimized for use with smartphones. As these keyboard layouts are not commonly used, switching to them results in a poor novice experience with an initially decreased text input speed. Previous investigations confirmed that this poor novice experience can be attenuated by using visual clues accentuating the most probable keys. This gives rise to the question of which visual impression leads to the best performance improvement. Therefore, this article evaluates the performance of six visual clues with different visual impressions via a user study conducted with 28 participants. The results showed that the visual clue with a pop-up animation and the visual clue with an increased font size performed the best with a 4% and 3%, respectively, typing performance improvement. Nonetheless, the questionnaire and personal preferences part of the user study showed that a 4% performance gain is not enough to crown one individual visual clue as the best visual clue to be used with every individual. This leads to the conclusion that functionality to choose one personal preference is more important than focusing on one specific best clue.
- 18 August 2023
- Published by MDPI AG in Applied Sciences
- Vol. 13 (16) , 9391
Open AccessLet Me Help You: Improving the Novice Experience of High-Performance Keyboard Layouts with Visual Clues
Since the advent of smartphones, work tasks shift progressively towards mobile phones. The main hurdle of this progression is the substantially slower input speed of smartphones in comparison with physical keyboards. This input speed could be improved by using keyboard layouts optimized for use with smartphones. As these keyboard layouts are not commonly used, switching to them results in a poor novice experience with an initially decreased text input speed. Previous investigations confirmed that this poor novice experience can be attenuated by using visual clues accentuating the most probable keys. This gives rise to the question of which visual impression leads to the best performance improvement. Therefore, this article evaluates the performance of six visual clues with different visual impressions via a user study conducted with 28 participants. The results showed that the visual clue with a pop-up animation and the visual clue with an increased font size performed the best with a 4% and 3%, respectively, typing performance improvement. Nonetheless, the questionnaire and personal preferences part of the user study showed that a 4% performance gain is not enough to crown one individual visual clue as the best visual clue to be used with every individual. This leads to the conclusion that functionality to choose one personal preference is more important than focusing on one specific best clue.
- Open Access
- 4 April 2023
- Published by MDPI AG in Sustainability
- Vol. 15 (7) , 6230
Cost Efficiency Evaluation of an On-Chain, Decentralized Ride-Sharing Platform
Traditional taxi providers are slowly losing their clients to online taxi providers, which are becoming increasingly popular. These companies typically provide a commercial, centralized platform in order to provide these services. The concept of centralization gives those providers significant power over their users, such as demanding and changing fees and collecting data over provided services. We aim to discuss how a decentralized approach using blockchains can solve those problems. A ride-sharing platform with the most essential functionality is implemented using a decentralized application based on smart contracts, which are, in turn, deployed on different blockchains. Various effects with respect to implementing and providing a decentralized ride-sharing service are taken in to account, especially regarding the effective cost for users. The fees of the developed platform heavily depend on which blockchain it is deployed on. Blockchains may compete in term of cost with traditional ride-sharing platforms, but choosing a suitable blockchain for the application is essential. Furthermore, the prototype shows that the time it takes for a transaction to be manifested into a block and subsequently confirmed is multiple times longer than a traditional web service response. These new problems require a new approach to implementing those requests that do not negatively affect the user experience.
- 4 April 2023
- Published by MDPI AG in Sustainability
- Vol. 15 (7) , 6230
Open AccessCost Efficiency Evaluation of an On-Chain, Decentralized Ride-Sharing Platform
Traditional taxi providers are slowly losing their clients to online taxi providers, which are becoming increasingly popular. These companies typically provide a commercial, centralized platform in order to provide these services. The concept of centralization gives those providers significant power over their users, such as demanding and changing fees and collecting data over provided services. We aim to discuss how a decentralized approach using blockchains can solve those problems. A ride-sharing platform with the most essential functionality is implemented using a decentralized application based on smart contracts, which are, in turn, deployed on different blockchains. Various effects with respect to implementing and providing a decentralized ride-sharing service are taken in to account, especially regarding the effective cost for users. The fees of the developed platform heavily depend on which blockchain it is deployed on. Blockchains may compete in term of cost with traditional ride-sharing platforms, but choosing a suitable blockchain for the application is essential. Furthermore, the prototype shows that the time it takes for a transaction to be manifested into a block and subsequently confirmed is multiple times longer than a traditional web service response. These new problems require a new approach to implementing those requests that do not negatively affect the user experience.
- Open Access
- 2 March 2022
- Published by MDPI AG in Applied Sciences
- Vol. 12 (5) , 2590
The Chef’s Choice: System for Allergen and Style Classification in Recipes
Allergens in food items can be dangerous for individuals affected by food allergens. Considering how many different ingredients and food items exists, it is hard to keep track of which food items contain relevant allergens. Food businesses in the EU are required to label foods with information about the 14 major food allergens defined by the EU legislation. This improves the situation for affected individuals. Nevertheless, more changes are necessary to provide reasonable protection for people with severe allergic reactions. Recipe websites and online content is usually not labelled with allergens. In addition, the 14 main allergen categories consist of a variety of different ingredients that are not always easy to remember. Scanning websites and recipes for specific allergens can consume a fair amount of time if the reader wants to make sure no allergen is missed. In this article, a dataset is processed and used for machine learning to classify cuisine style and allergens. The dataset used contains labelling for the 14 major allergen categories. Furthermore, a system is proposed that informs the user about style and allergens in a recipe with the help of a browser add-on. To measure the performance of the proposed system, a user study is conducted where participants label recipes with food allergens. A comparison between human and system performance as well as the time needed to read and label recipes concludes this article.
- 2 March 2022
- Published by MDPI AG in Applied Sciences
- Vol. 12 (5) , 2590
Open AccessThe Chef’s Choice: System for Allergen and Style Classification in Recipes
Allergens in food items can be dangerous for individuals affected by food allergens. Considering how many different ingredients and food items exists, it is hard to keep track of which food items contain relevant allergens. Food businesses in the EU are required to label foods with information about the 14 major food allergens defined by the EU legislation. This improves the situation for affected individuals. Nevertheless, more changes are necessary to provide reasonable protection for people with severe allergic reactions. Recipe websites and online content is usually not labelled with allergens. In addition, the 14 main allergen categories consist of a variety of different ingredients that are not always easy to remember. Scanning websites and recipes for specific allergens can consume a fair amount of time if the reader wants to make sure no allergen is missed. In this article, a dataset is processed and used for machine learning to classify cuisine style and allergens. The dataset used contains labelling for the 14 major allergen categories. Furthermore, a system is proposed that informs the user about style and allergens in a recipe with the help of a browser add-on. To measure the performance of the proposed system, a user study is conducted where participants label recipes with food allergens. A comparison between human and system performance as well as the time needed to read and label recipes concludes this article.
- Open Access
- 10 December 2021
- Published by MDPI AG in Applied Sciences
- Vol. 11 (24) , 11756
Continuous Mobile User Authentication Using Combined Biometric Traits
Current state of the art authentication systems for mobile devices primarily rely on single point of entry authentication which imposes several flaws. For example, an attacker obtaining an unlocked device can potentially use and exploit it until the screen gets locked again. With continuous mobile user authentication, a system is embedded into the mobile devices, which continuously monitors biometric features of the person using the device, to validate if those monitored inputs match and therefore were made by the previously authenticated user. We start by giving an introduction towards the state of the art of currently used authentication systems and address related problems. For our main contribution we then propose, implement and discuss a continuous user authentication system for the Android ecosystem, which continuously monitors and records touch, accelerometer and timestamp data, and run experiments to gather data from multiple subjects. After feature extraction and normalization, a Hidden Markov Model is employed using an unsupervised learning approach as classifier and integrated into the Android application for further system evaluation and experimentation. The final model achieves an Area Under Curve of up to 100% while maintaining an Equal Error Rate of 1.34%. This is done by combining position and accelerometer data using gestures with at least 50 data points and averaging the prediction result of 25 consecutive gestures.
- 10 December 2021
- Published by MDPI AG in Applied Sciences
- Vol. 11 (24) , 11756
Open AccessContinuous Mobile User Authentication Using Combined Biometric Traits
Current state of the art authentication systems for mobile devices primarily rely on single point of entry authentication which imposes several flaws. For example, an attacker obtaining an unlocked device can potentially use and exploit it until the screen gets locked again. With continuous mobile user authentication, a system is embedded into the mobile devices, which continuously monitors biometric features of the person using the device, to validate if those monitored inputs match and therefore were made by the previously authenticated user. We start by giving an introduction towards the state of the art of currently used authentication systems and address related problems. For our main contribution we then propose, implement and discuss a continuous user authentication system for the Android ecosystem, which continuously monitors and records touch, accelerometer and timestamp data, and run experiments to gather data from multiple subjects. After feature extraction and normalization, a Hidden Markov Model is employed using an unsupervised learning approach as classifier and integrated into the Android application for further system evaluation and experimentation. The final model achieves an Area Under Curve of up to 100% while maintaining an Equal Error Rate of 1.34%. This is done by combining position and accelerometer data using gestures with at least 50 data points and averaging the prediction result of 25 consecutive gestures.
- Open Access
- 13 September 2021
- Published by MDPI AG in Applied Sciences
- Vol. 11 (18) , 8473
Comparing Human Activity Recognition Models Based on Complexity and Resource Usage
Human Activity Recognition (HAR) is a field with many contrasting application domains, from medical applications to ambient assisted living and sports applications. With ever-changing use cases and devices also comes a need for newer and better HAR approaches. Machine learning has long been one of the predominant techniques to recognize activities from extracted features. With the advent of deep learning techniques that push state of the art results in many different domains like natural language processing or computer vision, researchers have also started to build deep neural nets for HAR. With this increase in complexity, there also comes a necessity to compare the newer approaches to the previous state of the art algorithms. Not everything that is new is also better. Therefore, this paper aims to compare typical machine learning models like a Random Forest (RF) or a Support Vector Machine (SVM) to two commonly used deep neural net architectures, Convolutional Neural Nets (CNNs) and Recurrent Neural Nets (RNNs). Not only in regards to performance but also in regards to the complexity of the models. We measure complexity as the memory consumption, the mean prediction time and the number of trainable parameters of the models. To achieve comparable results, the models are all tested on the same publicly available dataset, the UCI HAR Smartphone dataset. With this combination of prediction performance and model complexity, we look for the models achieving the best possible performance/complexity tradeoff and therefore being the most favourable to be used in an application. According to our findings, the best model for a strictly memory limited use case is the Random Forest with an F1-Score of 88.34%, memory consumption of only 0.1 MB and mean prediction time of 0.22 ms. The overall best model in terms of complexity and performance is the SVM with a linear kernel with an F1-Score of 95.62%, memory consumption of 2 MB and a mean prediction time of 0.47 ms. The two deep neural nets are on par in terms of performance, but their increased complexity makes them less favourable to be used.
- 13 September 2021
- Published by MDPI AG in Applied Sciences
- Vol. 11 (18) , 8473
Open AccessComparing Human Activity Recognition Models Based on Complexity and Resource Usage
Human Activity Recognition (HAR) is a field with many contrasting application domains, from medical applications to ambient assisted living and sports applications. With ever-changing use cases and devices also comes a need for newer and better HAR approaches. Machine learning has long been one of the predominant techniques to recognize activities from extracted features. With the advent of deep learning techniques that push state of the art results in many different domains like natural language processing or computer vision, researchers have also started to build deep neural nets for HAR. With this increase in complexity, there also comes a necessity to compare the newer approaches to the previous state of the art algorithms. Not everything that is new is also better. Therefore, this paper aims to compare typical machine learning models like a Random Forest (RF) or a Support Vector Machine (SVM) to two commonly used deep neural net architectures, Convolutional Neural Nets (CNNs) and Recurrent Neural Nets (RNNs). Not only in regards to performance but also in regards to the complexity of the models. We measure complexity as the memory consumption, the mean prediction time and the number of trainable parameters of the models. To achieve comparable results, the models are all tested on the same publicly available dataset, the UCI HAR Smartphone dataset. With this combination of prediction performance and model complexity, we look for the models achieving the best possible performance/complexity tradeoff and therefore being the most favourable to be used in an application. According to our findings, the best model for a strictly memory limited use case is the Random Forest with an F1-Score of 88.34%, memory consumption of only 0.1 MB and mean prediction time of 0.22 ms. The overall best model in terms of complexity and performance is the SVM with a linear kernel with an F1-Score of 95.62%, memory consumption of 2 MB and a mean prediction time of 0.47 ms. The two deep neural nets are on par in terms of performance, but their increased complexity makes them less favourable to be used.