Mon ÉTS  |  Bibliothèque  |  Bottins  |  Emplois  |  Intranet  |  ENGLISH

Offre de projet de recherche

Audio processing algorithm for digital audio filtering of a smart earplug for the operating room and intensive care unit

Programme d'études visé
Maîtrise avec mémoire   

Connaissances requises
Étudiant(e) intéréssé par le traitement de signal et les mesures expérimentales pour des applications biomédicales   
Début du projet : 2018-01-01

Domaines de recherche
Génie Électrique,
Génie Logiciel,
Génie Mécanique,
Génie de la production automatisée,
Génie des Technologies de l'information,
Génie, concentration Réseaux de télécommunications,
Génie, concentration en Technologies de la santé

Description
Audio processing algorithm for digital audio filtering of a smart earplug for the operating room and intensive care unit.

Sound exposure in the hospital can have deleterious effects on patients and practitioners. Clinicians perform worse on tasks involving patient monitoring in noisy and highly attentionally demanding environments. Research on the signal-to-noise ratio of alarms can decrease the overall sound exposure by decreasing the alarm fraction contribution of total sound.

Alarms in the ICU sound frequently and 85-99% of cases do not require clinical intervention. As alarm frequency increases, clinicians develop ‘alarm fatigue’ resulting in desensitization, missed alarms, and delayed responses. This is dangerous for the patient when an alarm-provoking event requires clinical intervention but is inadvertently missed. Alarm fatigue can also cause clinicians to: set alarm parameters outside effective ranges to decrease alarm occurrence, decrease alarm volumes to an inaudible level; silence frequently insignificant alarms; and be unable to distinguish alarm urgency. Since false alarm and clinically insignificant alarm rates reach 80-90%, practitioners distrust alarms, lose confidence in their significance, and manifest alarm fatigue.

Yet, failure to respond to the infrequent clinically significant alarm may lead to poor patient outcomes. Fatigue from alarm amplitude can be address pending that these alarms were produced at proper signal-to-noise ratios (SNR), as recent studies demonstrated that low to moderate SNR where sufficient for clinicians to react properly to alarms while considerably decreasing their perceived levels and annoyance.

The goal of this research project is to develop an real-time audio processing algorithm that could be implemented on a digital hearing protector equipped with external microphone and internal loudspeaker, such as the one developed by the EERS-ETS Industrial Research Chair in In-Ear Technologies (CRITIAS). To achieve this goal, the following steps will be conducted by an intern student, in collaboration with Dr. Schlesinger, and Prof. Voix:

1.Manual segmentation of real-world audio recordings of ICU rooms (provided by Prof. Schlesinger), in order to identify quiet periods, alarm time epochs, speech activities, and co-occurrence of the aforementioned to determine masking effects;

2. Adaptation of an automated alarm detection algorithm, previously developed in MATLAB by CRITIAS (provided by Prof. Voix), for the detection of alarm time epochs;

3.Development of a real-time processing audio algorithm based on spectral substraction, utilizing the alarm period previously identified in step 2, and dynamically adjusting the SNR of the auditory alarm signals to match the target SNR for a given background noise level informed by work completed by Dr. Schlesinger;

4.Validation of the developed algorithm off-line on real-world audio recordings for proper SNR adjustment;

5. Implementation of the existing alarm detection algorithm (step 2) and newly developed denoising and SNR
adjustment algorithms on the Auditory Research Platform (ARP) provided by CRITIAS;

6.Validation of novel smart earplug during clinical scenarios utilizing the Center for Experiential Learning and Assessment (CELA) high-fidelity medical simulation center at Vanderbilt University Medical Center.



Financement
Financement assuré via la chaire de recherche CRITIAS et/ou par projet MITACS   

Autres informations
Partenaires Impliqués
EERS Technologies Inc (eers.ca)
Vanderbilt University Medical Center (TN, USA)
NSERC-EERS Industrial Research Chair in In-Ear Technologies (critias.etsmtl.ca)

Responsable à contacter

Professeur, Voix Jérémie Département de génie mécanique
Téléphone : 514 396-8437
Télécopieur 514 396-8530
Courriel jeremie.voix@etsmtl.ca