A Dimensional Emotion Recognition Method Based on Multi-scale Time Series Modeling

Abstract

The invention discloses a dimensional emotion recognition method based on multi-scale time series modeling. The method includes the following steps: performing face detection and tracking for each frame of images in a video sequence, and extracting face key points as a first type group Face features; extract the gray values ​​of the pixels in the face area image, the face mouth area image and the face eye area image as the second, third and fourth groups of face features; The four categories of facial features are used to perform preliminary dimensional emotion prediction; according to the initial emotion prediction results of consecutive N unit time periods t, the linear regression is used to perform time series and modal fusion, and output the emotional prediction value of the video sequence. The method of the invention performs time sequence modeling of different scales on the video sequence signal, and realizes accurate prediction of each time sequence unit in the sequence. The invention is suitable for emotion recognition of face signals in videos, and has the advantages of good real-time performance, and can greatly improve recognition accuracy.

Description

technical field [0001] The invention belongs to the field of video signal processing, and in particular relates to a method for dimensional emotion recognition based on multi-scale time series modeling, which improves the accuracy of continuous dimensional emotion recognition. Background technique [0002] In recent years, researchers at home and abroad have done a lot of research work on continuous dimension emotion recognition, and proposed many effective methods for emotion recognition. These methods can be divided into detection methods based on static classifiers and detection methods based on dynamic classifiers in terms of processing strategies. Detection methods based on static classifiers mostly use support vector machines (SVM), neural networks, Boosting, etc., and most of these classifiers are discriminative models. Due to its strong discrimination ability, it is widely used in the field of emotional state recognition, but this method ignores the fact that emotio...

AI Technical Summary

Problems solved by technology

However, HMM can only perform temporal modeling on a single time scale, and the scope of temporal modeling is limited, which cannot fully reflect the key role of emotional temporal information in emotion recognition.

Images