Développement d’une méthode de mesure des biosignaux acoustiques dans l’oreille occluse

Abstract

In heavy industries and confined spaces, workers are prone to develop noise induced hearing loss (NIHL) and to have work-related accidents and injuries. Workers already wear hearing protection devices (HPD) in order to prevent NIHL, but HPDs do not monitor workers’ health in order to prevent or detect a sudden illness or injury. Continuous health monitoring is an essential feature to detect a sudden change in vital signs such as heart rate or breathing rate. So far, many wearable devices have been developed in order to monitor heart rate or breathing rate : watches, shirts, wrist bands or belts. However, those wearables are not appropriate for workers health monitoring because they already wear many personal protection equipments. This project aims to integrate a non-invasive health monitoring feature into the workers’ HPDs. The proposed method takes advantage of audio hardware already integrated in their HPDs without adding hardware, which leads to a significant cost saving. Heartbeat and respiration sounds were measured with a microphone positioned within the ear canal, which already picks up sounds for communication, smart noise protection and in-ear dosimetry. To develop heart and breathing rates extraction algorithms, a unique 25-person database was created from sounds measured in the occluded ear canal. Subjects were asked to breathe at various rhythms and intensities through their mouth or nose and these real-life sounds were recorded. Heart and respiration sound features were investigated when recorded at this specific location. Subsequent algorithms were developed in order to assess the user’s heart and breathing rates. Results from the algorithms were then compared to the numerical values obtained by a reference device used during the measurement. Then, noise was added to the recorded signal and a denoising algorithm was applied to test the robustness of the extraction algorithms in noisy environments. Developed algorithms analyzed 20 subjects and 12 hours of recordings. The overall averaged Least Absolute Deviation (aLAD) for the heart rate and breathing rate extraction were respectively, 4.3 beats per minute and 3.0 cycles per minute. Simulation results indicated that the maximum environmental noise for the heart rate extraction was 110 dB, whereas the extraction of breathing rate with noise lacked of accuracy. This proof of concept enables the development of in-ear technologies for a wide range of non-invasive health and safety monitoring, ranging from industrial workers monitoring through their hearing protection devices to the watching of vital signs of elderly people through their hearing aids.

Date	16 Nov 2016
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Jérémie Voix (Supervisor)