Snore recognition using a reduced set of spectral features

Snoring affects the sleep quality of the snorer itself and its social circle. Some types of snoring are related to sleep apnea, which leads to sleepiness during the day and to several health risks. Thus automatic detection of the different types of snoring may lead to more specific diagnosis and consequent treatment. In this work, we propose to use a reduced set of speech related features that includes spectral information, Mel-Frequency Cepstral Coefficients (MFCCs), prosodic values and bioinspired information. Extreme Learning Machines (ELM) are proposed to learn on the non-linear feature set. A well-known classifier as Support Vector Machines (SVM) is used as baseline. Several configurations for the feature sets and the ELM were evaluated. The bioinspired information shows promising results on the Munich-Passau Snore Sound Corpus (MPSSC) with respect to the baseline performance on the development partition.