Recognition of manual forest work for time on task recording

In the manual forestry, the worker carries his equipment to a tree by foot. There the tree is felt and processed. Depending on the surrounding, this needs more or less time. This paper automatically analyses the needed time for different activities. Therefore, the worker gets an estimate of his time spends on different tasks and can estimate his productivity. The approach uses therefor a mobile phone and a smartwatch. This work could also be used to conduct general time studies for time and equipment comparison. The focus is on the frequency analysis, feature creation and the obstacles, such as asynchrony of the recordings and labeling errors during the annotation of the ground truth, which are described and calculated.