Respiratory frequency estimation from accelerometric signals acquired by mobile phone in a controlled breathing protocol

The aim of this work was to test if the smartphone's embedded triaxial accelerometer can be used to extract respiratory frequency information from the chest movements during a controlled breathing protocol. Respiratory signals from 10 young volunteers were recorded simultaneously, by two smartphones (iPhone 4s and 6s; sampling frequency ~100 Hz), positioned one on the sternum and one on the belly, while in supine posture. At the same time, a belt transducer was used to acquire the reference respiratory signal. A controlled breathing protocol, consisting of four consecutive phases of 12 respiratory cycles each (respiratory frequencies at 0.25, 0.17, 0.125 and 0.1 Hz), was imposed through the visualization of a moving bar on a display. After low-pass filtering (fc=0.5 Hz), the respiratory signal was obtained from both smartphones, and respiratory frequency derived for each phase. Compared to the belt transducer, the resulting error was lower than 2% for each imposed respiratory frequency, for both smartphones' positions, with better results obtained for the smartphone positioned above the belly.