Obs tructive sleep apnea (OSA) is a highly prevalent chronic disease, especially in elderly and obese population . Despite constitut ing a huge health and economic problem, most patients remain undiagnosed due to limitations in current strategies. Therefore, it is essential to find cost - effective diagnostic alternatives. One of these novel approaches is the analysis of acoustic snoring signals. Snoring is an early symptom of OSA which carr ies pathophysiological information of high diagnostic value. For this reason, the main objective of this work is to study the characteristics of single snores of different types, from healthy and OSA subjects. To do that, we analyzed snoring signals from p revious databases and developed an experimental protocol to record simulated OSA - related sounds and characterize the response of two commercial tracheal microphones. Automatic programs for filtering, downsampling, event detection and time - frequency analysi s were built in MATLAB. We found that time - frequency maps and spectral parameters (central, mean and peak frequency and energy in the 100 - 500 Hz band) allow distinguishing regular snores of healthy subjects from non - regular snores and snores of OSA subject s. Regarding the two commercial microphones, we f ound that one of them was a suitable snoring sensor, while the other had a too restricted frequency response. Future work shall include a higher number of episode s and subjects , but our study has contributed to show how important the differences between regular and non - regular snores can be for OSA diagnosis, and how much clinically relevant information can be extracted from time - frequency maps and spectral parameters of single snores

Peer Reviewed