SCarrie: A Real-Time System for Sound Event Detection for Assisted Living
Marko Chavdar, Tomislav Kartalov, Zoran A. Ivanovski, Dimitar Taškovski, Branislav Gerazov · 2023
Science aims to improve the quality of human life by making it easier and more comfortable. As the human life span increases, new assisted living technologies are required in order to keep living standards high. We developed a real-time system for remote acoustic monitoring in assisted living scenarios. The system is based on a model for acoustic event classification using Convolutional Neural Networks. To train and evaluate the system, we designed and created a dataset that comprises relevant sound events as well as background noise. To optimize the system we evaluated different acoustic features that have been found to work well in speech recognition. We deployed the system on a small single-board computer and the results show that the system has a satisfactory level of performance and it is well suited for the task.