Domain adaptive behavior recognition system and method based on data enhancement and pseudo tag

Abstract

The invention discloses a domain adaptive behavior recognition system and method based on data enhancement and a pseudo tag, and the system comprises a data collection module which collects the original data of a sensor when a user moves; the data preprocessing module is used for carrying out normalization and data segmentation on the source domain sensor data in the public data set and the collected target domain sensor data to obtain source domain and target domain sample data; the behavior recognition module is used for selecting source domain sample data for training to obtain a domain generalization basic model, performing data enhancement on target domain sample data to obtain enhanced target domain sample data, and optimizing the domain generalization basic model by using an iterative calculation pseudo-label algorithm and the enhanced target domain sample data to obtain the behavior recognition result. The classification result of the target domain sample data when iteration is terminated is the behavior recognition result. According to the method, the classification model is optimized by using the unmarked target domain sample data, the dependence on the marked sample of the target domain is reduced, and the accuracy and robustness of behavior recognition are improved while the marking cost is reduced.

Description

technical field [0001] The invention belongs to the fields of intelligent perception technology and artificial intelligence technology, and in particular relates to a domain adaptive behavior recognition system and method based on data enhancement and pseudo-labels. Background technique [0002] At present, mobile smart devices and wearable smart devices have become an important part of people's daily life, and people use these devices to detect and record their behavioral states. Sensors compatible with mobile devices and wearable devices save and record data generated by user activities, which are widely used in behavior recognition tasks, such as motion detection, human-computer interaction, and other applications. [0003] For behavior recognition tasks based on wearable devices, the traditional solution is to train the behavior recognition model in a supervised manner, that is, to train the classification model through a large number of labeled sample data, and then use...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the deficiencies of the above-mentioned prior art, the purpose of the present invention is to provide a domain adaptive behavior recognition system and method based on data enhancement and pseudo-labels to solve the problem of existing behavior recognition models in new data domains without labeled data ( That is, the problem of low performance when performing behavior recognition tasks on the target domain); the invention can make the model use the unlabeled data on the target domain to adjust and optimize, reduce the dependence on the labeled samples in the target domain, and improve the efficiency while reducing the marking cost. Accuracy and Robustness of Action Recognition

Images