Abstract: To quantify the firefighter's working efficiency accurately and protect the firefighter's life safety in the fire rescue environment, the study improves the algorithm through a posterior probabilistic transformation of the Support Vector Machine (SVM), and proposes a working efficiency evaluation method based on the fusion of the SVM and the Dempster-Shafer(DS). First, the physiological and psychological parameters of firefighters' characteristics are extracted; Second, feature parameters are classified and predicted by SVM regression; Then, the predicted results are transformed into the basic probability assignments of DS evidence by a posteriori probability transformation; Finally, firefighter's working efficiency is evaluated in real time according to the DS evidence theory. The results show that the mean square error of SVM regression for predicting the heart rate of working efficiency parameter is 0.002, and the coefficient of determination is 0.95, which is better than that of BP neural network. The average absolute error of SVM-DS fusion algorithm for evaluation of working efficiency is 10.65%, and it can be better realized to evaluate firefighters' working efficiency in real time. The algorithm presented in this study can provide theoretical data for developing early warning system, determining the maximum safe working time, and realizing scientific rescue and enhancement of disaster response capability by effectively quantifying the working efficiency and evaluating it in real-time.

Key words: working efficiency evaluation, firefighters, multi-feature fusion, support vector machines, DS evidence theory