METHOD FOR RECOGNITION AND CLASSIFICATION OF ACTIONS OF MANUAL LABOR WORKERS BY DATA FROM WEARABLE DEVICES EQUIPPED WITH PHYSICAL ACTIVITY SENSORS Russian patent published in 2022 - IPC G06Q10/06 G06N20/00 

FIELD: labor rationing.

SUBSTANCE: invention relates to the field of manual labor rationing; it can be used for building an activity profile and a list of actions performed by manual labor workers during working hours for subsequent analysis and determination of labor efficiency, as well as optimization of labor processes based on processing of time series (signals) of readings of physical activity sensors fixed to the human body. A method for recognition and classification of actions of manual labor workers by data from wearable devices equipped with physical activity sensors is characterized by: collection of data from physical activity sensors, analysis of types of actions of manual labor workers, when performing a type of labor activity under consideration; formation of a hierarchical classification model; formation of a training data sampling; building an algorithmic composition above a combination of built classification algorithms for each and/or separate levels of hierarchy of classes, which are used in the future for automatic marking of new data.

EFFECT: possibility of classification of manual labor of workers during the performance of work duties, or classification of other complex/composite types of human physical activity with unexpressed periodic character of signals or time series of readings of physical activity sensors.

1 cl