Roobesh Balaji
Roobesh Balaji
Advisors
Rebecca Lennartz (M. Sc.), Maike Stoeve (M. Sc.) Prof. Dr. Björn Eskofier
Duration
11 / 2023 – 05 / 2024
Abstract
Understanding how and why players perform at their best is crucial to improving physical education and training procedures. While athletic performance is quantifiable and can be objectively measured, the underlying factors that influence performance remain less understood. Motivation is discussed as a critical factor of a player’s performance, being influential in initiating, directing, sustaining, and terminating efforts by the player [1]. Methods of understanding motivation are questionnaires such as the Intrinsic Motivation Inventory (IMI), but these questionnaires tend to have limitations as they are not deployable to measure motivation continuously and objectively during the sporting drill. Therefore, this thesis investigates approaches to indirectly assess motivation during sports based on athletes’ performance.
An increase in computational efficiency and improved deep learning algorithms have led to video-based modality being a crucial tool used in sports performance analysis, action recognition, and event detection [2] [3] [4]. Combining more than one modality can reveal patterns and correlations that can be beneficial for sports analytics as it can provide more comprehensive and nuanced insights that would be difficult to gain from a single source [5]. Zhang et al. show that the utilization of IMU data in conjunction with video-based modalities improves information extraction through multimodal approaches [6]. Jones et al. discuss that gamification can be used to enhance motivation. Therefore, in this thesis, video and IMU data are fused to analyze the performance of athletes while performing a drill.
A soccer rondo drill in a 360-degree environment is taken as an example scenario in which various gamification elements can be used to influence player motivation, as shown in the literature [7]. In this work, gamification encompasses elements such as a team leaderboard, audio cues, achievable badges with progress bars, and a point counter. To investigate the influence of these gamification elements on motivation and performance, the player is observed in two distinct scenarios while performing the same drill. One scenario contains gamification elements while the other scenario does not. This work aims to develop a framework to determine the effectiveness of multimodal algorithms in classifying between the two scenarios.
To conduct a comprehensive analysis, this exploratory approach will look into multimodal data fusion methods using overhead video data and IMU data from chest and head to assess player performance during the drills [8]. Preprocessing techniques will be employed to synchronize and normalize data from different sources, with a primary focus on extracting features as quantifiable performance indicators including speed, movement patterns, and response time of the player. To observe the impact of gamification elements, this work will statistically evaluate the metrics from the features that were extracted. Furthermore, these performance-based features will be used to classify between the two scenarios. Motivational questionnaires collected after each scenario will be used to provide contextual insight into players’ motivational factors to evaluate the relationship between motivation and performance in a gamified soccer drill.
References
[1] Nur Aina Syuhada Omar, Razif Sazali, and Nur Mim Naimah Zainuddin. Understanding the motivation factors towards sports performance within the scope of collegiate athletes. Journal of Contemporary Social Science and Education Studies (JOCSSES), 3(1):73–79, 2023.
[2] Henry Friday Nweke, Ying Wah Teh, Mohammed Ali Al-Garadi, and Uzoma Rita Alo. Deep learning algorithms for human activity recognition using mobile and wearable sensor networks: State of the art and research challenges. Expert Systems with Applications, 105:233–261, 2018.
[3] Preksha Pareek and Ankit Thakkar. A survey on video-based human action recognition: recent updates, datasets, challenges, and applications. Artificial Intelligence Review, 54:2259–2322, 2021.
[4] Hongji Guo, Alexander Aved, Collen Roller, Erika Ardiles-Cruz, and Qiang Ji. Video-based complex human event recognition with a probabilistic transformer. In Geospatial Informatics XIII, volume 12525, pages 184–192. SPIE, 2023.
[5] Serena Lee-Cultura, Kshitij Sharma, Sofia Papavlasopoulou, and Michail Giannakos. Motionbased educational games: Using multi-modal data to predict player’s performance. In 2020 IEEE conference on games (cog), pages 17–24. IEEE, 2020.
[6] Cheng Zhang, Tianqi Zu, Yibin Hou, Jian He, Shengqi Yang, and Ruihai Dong. Human activity recognition based on multi-modal fusion. CCF Transactions on Pervasive Computing and Interaction, pages 1–12, 2023.
[7] Matthew Jones, Jedediah E Blanton, and Rachel E Williams. Science to practice: Does gamification enhance intrinsic motivation? Active Learning in Higher Education, 24(3):273–289, 2023.
[8] Jing Gao, Peng Li, Zhikui Chen, and Jianing Zhang. A survey on deep learning for multimodal data fusion. Neural Computation, 32(5):829–864, 2020.