Health Monitoring for Firefighters

Type: Master thesis
Student: Philip Dieringer
Adviser:  Burcu Cinaz, Bert Arnrich, Gerhard Tröster

Employing a sensor system which is capable of measuring physiological and activity data for firefighting tasks was the aim of this thesis. In order to investigate the influence of cognitive workload, an experiment design was presented to induce and discriminate three different levels of cognitive workload. According to the subjective ratings from NASA-TLX questionnaires and the performance scores obtained from N-Back-Tests, we could show, that all participants perceived the induced workload as intended from the experiment. As objective measures, the following parameters were recorded: heart rate, breathing rate and salvia samples. From heart rate a total of 23 features were calculated. A reduction of the feature subset was achieved in order to avoid redundant features. In order to discriminate the different workload levels various classifiers were compared as well as different cross-validation methods were implemented. The kNN-classifier yielded the best accuracy of 87.21% in discriminating the three defined cognitive workload levels.
In a second experiment, the influence of cognitive load on the gait behaviour was investigated. By implementing a heel strike algorithm, the variability of step durations has been analysed. It could be shown that the variability of step durations decreased with addition of cognitive workload.