Continuous learning-based user expression recognition method and system

Abstract

The invention discloses a continuous learning-based user expression recognition method and system. The method comprises a data acquisition step, a continuous learning step and a user expression test step. The method comprises the following steps: firstly, collecting different types of user expressions including feature data and feature mark data; training a model by using the expression type of the current user; establishing a mapping relation between previous expressions and new expressions by adopting an optimal transport technology when the new expressions of the user are learned, so as to reuse a previous model; and finally, updating the existing model by using the newly collected user expressions. The invention can solve the problems that new and old expressions of the user cannot be obtained at the same time and a new model can only be established by using the existing model and the new expression of the user, and the method occupies few resources in the implementation process and is high in applicability.

Description

technical field [0001] The present invention relates to an expression recognition method under the condition that the type of user expression changes with time evolution in an open dynamic environment, in particular to a method and system for user expression recognition across mobile devices based on continuous learning. Background technique [0002] Mobile device applications often rely on user expression recognition. For example, for applications that rely on user emotions, products and services can be adaptively recommended for users according to their different expression states, so as to improve the success rate of recommendation and improve user experience. However, the display environment is mostly open and dynamic, and the user expression data collected at different times tends to have large differences, and the types of user expressions tend to increase over time, becoming more fine-grained and divisible. Therefore, only new types of user expression data can be coll...

AI Technical Summary

Problems solved by technology

Therefore, the forgotten model reuse technology is mostly limited to the application scenarios of model adaptation data, and is not suitable for open and dynamic environments where user expression categories change, especially in application scenarios where the types of user expressions on mobile devices increase.

Images