Practical evaluation of the viability of sensorsand algorithms to detect falling and burial by avalanche
Justus Conradi, Patrik Tiainen · KTH Publication Database DiVA (KTH Royal Institute of Technology) · 2021
Avalanches kill on average 250 people per year. Fall detection algorithms have been used to detect many different types of falls. In this work, a robust humanoid test rigwas designed and built to investigate the use of proven threshold based fall detection algorithms during skiing. A novel burial detection algorithm is used, to detect whether or not a person has been buried by snow, as this is often lethal. Data acquisition were conducted in the Swedish mountains, as well as computer simulations to further strengthen the results. This thesis shows that an IMU, in conjunction with force sensors around the chest and shoulders are sufficient to detect a fall and burial by snow. The algorithms achieve excellent results based on measurement data gathered by thetest rig. Tests were made to check the algorithm for a person gathering data. This showed that additional evaluation is needed to be applicable to live people, since skiing by the test rig is not completely representative of human skiing.