Développement d'un capteur respiratoire non intrusif

During this project, a new model for recording, processing and storing respiratory signals was
developed, based on an existing prototype. The goal of the internship was divided into 3 main parts:
improve the design of the model, improve data acquisition and processing, and to compare the model’s
results with the industry standard for validation.
The methods and tools used to complete the project are based on the principles and concepts of
systems engineering, a discipline that processes complex projects as systems and provides an iterative
framework to methodically work towards solutions.
First the goals of the project were precisely formulated. Then an analysis of the state of the art was
carried out to identify possible weaknesses and limits of the current model, after which requirements
were defined. Based on the requirements possible solutions were considered and solution selection
was done through the use of a cost-benefit analysis.
Two concepts were designed and implemented as two iterations of the model, since the first model
showed unanticipated weaknesses during the first test phase that were incompatible with the system’s
requirements. Each iteration consists of a 3D and electrical design, a microprocessor routine and a data
handling process on the computer. This includes designing and printing a 3D model of the sensor
housing, designing and building the electronic parts of the system, programming the microcontroller to
acquire and filter the data correctly, acquire the data on the computer and visualize a live feed to
control signal strength.
Each iteration went through a conception phase, an implementation phase and a validation phase. After
the second iteration satisfied the requirements during functional testing, the system was used in
real-world conditions in the simulator for validation in multiple testing sessions at the HumanTech lab.
Results were analyzed with a python module called neurokit, with a script written for that purpose. A
second script was written to parse and display the results, and a Matlab function was developed to
perform post-acquisition filtering on recorded data, to understand the most efficient filters for
respiratory data.
This work showed the possibilities and limitations of an unobtrusive respiratory monitoring system for
use inside of vehicles. Error margins obtained were consistent, and improvements suggested.