Online adaptation method based on pattern matching and mobile perception scene

Abstract

The invention discloses an online adaptation method based on pattern matching and a mobile perception scene. The method comprises the following steps: carrying out classification identification by comparing the similarity between a target sample and a template in a template library to obtain a classification result; according to the obtained classification result, feedback from the application or the user is monitored in real time; in combination with a feedback result, the representativeness of the current target sample is judged according to an active learning theory, and the representativeness reflects the importance degree of the target sample for classification and recognition; and on the basis of the corresponding representativeness, updating the template library by a set strategy to realize online adaptation. By utilizing the method and the device, a developer can realize an application system supporting online continuous adaptation on the mobile equipment under the condition of ensuring the storage and calculation efficiency.

Description

technical field [0001] The present invention relates to the technical field of mobile perception, and more particularly, to an online adaptation method based on pattern matching and mobile perception scenarios. Background technique [0002] With the rapid development of micro-motor systems, smart devices such as mobile terminals have become part of people's daily life. As technology advances, more and more sensors can be integrated into a mobile device. Perceiving and providing services to users through mobile devices has also become a popular research direction. Human behavior recognition is one of the most important branches of human-centered mobile perception, and its common applications cover gesture recognition, daily behavior inference and other scenarios. However, most of the current recognition systems still have great difficulties when deploying to real scenes. Affected by various unknown condition changes in real scenes, the system recognition accuracy is signif...

AI Technical Summary

Problems solved by technology

However, most of the current recognition systems still have great difficulties when deployed to real scenes

Affected by changes in various unknown conditions in the real scene, the recognition accuracy of the system is significantly reduced

Especially for user-centric scenarios, individual user differences and dynamic uncertainties including usage habits and interest drift will bring huge challenges to the actual performance of the recognition system.

[0003] The existing recognition technology mainly starts with the training of the perception model, and minimizes the impact on the user experience by reducing the training overhead, but ignores the condition changes in the actual use process, resulting in poor recognition effect

Images