“Given audio recordings from an “in-ear” microphone, can we accurately detect occurrences of snacking events? Moreover, can we achieve this while respecting the computational, memory and power limitations of mobile devices? Researchers at the Aristotle University of Thessaloniki (AUTH) have developed algorithms that effectively address these issues, using extensive datasets recorded with chewing sensors developed by CSEM. More specifically, a fast, low-sampling rate algorithm was developed for detecting snacks based on the fractal dimension of the chewing signals. The algorithm achieved a precision of 87% and recall of 98% for snacking detection, surpassing the effectiveness of state-of-the-art methods in the same dataset. The results were presented in oral and poster presentations by Vasileios Papapanagiotou and Christos Diou at the 1st International Workshop on Multimedia Assisted Dietary Management (Madima), at Genova, Italy on 8th September, 2015.”
A second paper was presented by Christos Maramis of Aristotle University of Thessaloniki in the same conference. It is titled “Objective and Subjective Meal Registration via a Smartphone Application” and describes the mechanisms that have been designed and implemented in the SPLENDID Smartphone App for sensor-based and self-report meal registration, as well as for user feedback.
MADIMA2015 has been organized in the context of the 18th International Conference on Image Analysis and Processing (ICIAP2015).