Face-to-face view and sound multi-view emotion discrimination method and system

Abstract

The invention provides a face-to-face view and sound multi-view emotion discrimination method, and the method comprises the following steps: S1, obtaining initial video data, preprocessing the initialvideo data, and obtaining view data and audio data; S2, respectively extracting original features on the view data and the audio data; S3, performing secondary feature extraction to obtain potentialfeatures of the view data and the audio data; S4, fusing the potential features to obtain complete potential representation; and S5, classifying the complete potential representation into a pluralityof emotion categories with different probabilities, and outputting an emotion category with the highest probability as an emotion discrimination result. The provided face-to-face view and sound multi-view emotion discrimination system comprises a data preprocessing module, a feature extraction module, a degradation network module and a classification module. The invention face-to-face visual soundmulti-view emotion discrimination method and system solve the problem that the discrimination accuracy is not high enough because the existing emotion discrimination technology only depends on emotion discrimination of single-view data.

Description

technical field [0001] The present invention relates to the technical field of multi-view learning, and more specifically, relates to a face-to-face video-audio multi-view emotion discrimination method and system. Background technique [0002] With the gradual overcoming of core technologies in the field of computer software and hardware, the rapid development of the Internet industry has been promoted, and the advent of the era of big data has been accelerated. What follows is the exponential growth of data, making modern data increasingly complex and highly heterogeneous. The diversity of features of the same thing (feature acquisition means, feature processing methods, feature attributes, etc.) is very common in reality, and these features are considered to be multi-view data of the same category of objects. Different views in multi-view data are different reflections and descriptions of the same object, so different views have certain correlations. For example, a docto...

AI Technical Summary

Problems solved by technology

[0005] The present invention provides a face-to-face audio-visual multi-view emotion discrimination method and system in order to overcome the technical defect that the existing emotion discrimination technology only relies on single-view data for emotion discrimination and the discrimination accuracy is not high enough

Images