392254 Project: Head Gesture Recognition Using Hidden Markov Models (Pj) (WiSe 2026/2027)

Background and Motivation Recognizing head gestures such as nodding and shaking from video is fundamentally a sequence-classification problem. Hidden Markov Models (HMMs) provide a principled probabilistic framework for modeling temporal patterns in sequential data. In this project, an HMM-based classifier for head gesture recognition will be implemented to detect nods and shakes, following the approach of Tan et al. (2003). The model will be trained and evaluated on a manually annotated dataset, providing hands-on experience with probabilistic sequence modeling, temporal data analysis, and gesture recognition. Skills and Requirements Background in machine learning and probability. Familiarity with Python; experience with NumPy and scikit-learn. Basic understanding of probabilistic sequence models is advantageous. Strong analytical and problem-solving skills. Reference Tan, W., & Rong, G. (2003). A real-time head nod and shake detector using HMMs. Expert Systems with Applications, 25