Marc Kurz

Prof. Dr. Marc Kurz

University of Applied Sciences Upper Austria, Hagenberg Campus
orcid

Research Focus & Metrics

Academic Career Length

14Years (2013-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
13 September 2021 in Applied Sciences
Verified
Journal Article
Open access
PDF file
SASimon Angerbauer
Simon Angerbauer
APAlexander Palmanshofer
Alexander Palmanshofer
SSStephan Selinger
Stephan Selinger
Marc KurzMarc Kurz
Marc Kurz
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.

10 September 2020 in Applied Sciences
Verified
Journal Article
Open access
PDF file
Erik SonnleitnerErik Sonnleitner
Erik Sonnleitner
OBOliver Barth
Oliver Barth
APAlexander Palmanshofer
Alexander Palmanshofer
Marc KurzMarc Kurz
Marc Kurz
Since global road traffic is steadily increasing, the need for intelligent traffic management and observation systems is becoming an important and critical aspect of modern traffic analysis. In this paper, we cover the development and evaluation of a traffic measurement system for tracking, counting and classifying different vehicle types based on real-time input data from ordinary highway cameras by using a hybrid approach including computer vision and machine learning techniques. Moreover, due to the relatively low framerate of such cameras, we also present a prediction model to estimate driving paths based on previous detections. We evaluate the proposed system with respect to different real-life road situations including highway-, toll station- and bridge-cameras and manage to keep the error rate of lost vehicles under 10%.

29 July 2019 in Applied Sciences
Verified
Journal Article
Open access
PDF file
BDBirgitta Dresp-Langley
Birgitta Dresp-Langley
OEOle Kristian Ekseth
Ole Kristian Ekseth
JFJan Fesl
Jan Fesl
SGSeiichi Gohshi
Seiichi Gohshi
Marc KurzMarc Kurz
Marc Kurz
HSHans-Werner Sehring
Hans-Werner Sehring
Detecting quality in large unstructured datasets requires capacities far beyond the limits of human perception and communicability and, as a result, there is an emerging trend towards increasingly complex analytic solutions in data science to cope with this problem. This new trend towards analytic complexity represents a severe challenge for the principle of parsimony (Occam’s razor) in science. This review article combines insight from various domains such as physics, computational science, data engineering, and cognitive science to review the specific properties of big data. Problems for detecting data quality without losing the principle of parsimony are then highlighted on the basis of specific examples. Computational building block approaches for data clustering can help to deal with large unstructured datasets in minimized computation time, and meaning can be extracted rapidly from large sets of unstructured image or video data parsimoniously through relatively simple unsupervised machine learning algorithms. Why we still massively lack in expertise for exploiting big data wisely to extract relevant information for specific tasks, recognize patterns and generate new information, or simply store and further process large amounts of sensor data is then reviewed, and examples illustrating why we need subjective views and pragmatic methods to analyze big data contents are brought forward. The review concludes on how cultural differences between East and West are likely to affect the course of big data analytics, and the development of increasingly autonomous artificial intelligence (AI) aimed at coping with the big data deluge in the near future.

25 February 2021 in Applied Sciences
Verified
Journal Article
Open access
PDF file
Marc KurzMarc Kurz
Marc Kurz
RGRobert Gstoettner
Robert Gstoettner
Erik SonnleitnerErik Sonnleitner
Erik Sonnleitner
Since electronic components are constantly getting smaller and smaller, sensors and logic boards can be fitted into smaller enclosures. This miniaturization lead to the development of smart rings containing motion sensors. These sensors of smart rings can be used to recognize hand/finger gestures enabling natural interaction. Unlike vision-based systems, wearable systems do not require a special infrastructure to operate in. Smart rings are highly mobile and are able to communicate wirelessly with various devices. They could potentially be used as a touchless user interface for countless applications, possibly leading to new developments in many areas of computer science and human–computer interaction. Specifically, the accelerometer and gyroscope sensors of a custom-built smart ring and of a smartwatch are used to train multiple machine learning models. The accuracy of the models is compared to evaluate whether smart rings or smartwatches are better suited for gesture recognition tasks. All the real-time data processing to predict 12 different gesture classes is done on a smartphone, which communicates wirelessly with the smart ring and the smartwatch. The system achieves accuracy scores of up to 98.8%, utilizing different machine learning models. Each machine learning model is trained with multiple different feature vectors in order to find optimal features for the gesture recognition task. A minimum accuracy threshold of 92% was derived from related research, to prove that the proposed system is able to compete with state-of-the-art solutions.

2 March 2022 in Applied Sciences
Verified
Journal Article
Open access
PDF file
ARAndreas Florian Roither
Andreas Florian Roither
Marc KurzMarc Kurz
Marc Kurz
Erik SonnleitnerErik Sonnleitner
Erik Sonnleitner
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.

10 December 2021 in Applied Sciences
Verified
Journal Article
Open access
PDF file
DRDominik Reichinger
Dominik Reichinger
Erik SonnleitnerErik Sonnleitner
Erik Sonnleitner
Marc KurzMarc Kurz
Marc Kurz
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.