Signals and Telecommunication Journal Martin Frešer, Igor Košir, Violeta Mirchevska, Mitja Luštrek 8 2 2019 https://doi.org/10.6025/stj/2019/8/2/54-59 http://www.dline.info/stj/fulltext/v8n2/stjv8n2_2.pdf This paper proposes an elderly-care system, which uses a single sensing device installed in the user’s home, primarily based on a microphone. We present preliminary results on human activity recognition from sound data. The recognition is based on 19 types of sound features, such as spectral centroid, zero crossings, Melfrequency cepstrum coefficients (MFCC) and linear predictive coding (LPC). We distinguished between 6 classes: sleep, exercise, work, eating, home chores and home leisure. We evaluated the recognition accuracy using 4 supervised learning algorithms. The highest accuracy, obtained using support vector machines, was 76%.