Video emotion recognition method based on local enhanced motion history map and recursive convolutional neural network

Abstract

The invention discloses a video emotion recognition method based on a local enhanced motion history map and a recursive convolutional neural network, and the method comprises the steps: 1, obtaining astatic expression image data set and an expression video data set, and carrying out the preprocessing of the expression video data set; 2, calculating a local enhanced motion history map LEMHI; 3, pre-training a convolutional neural network VGG16 model by using the static picture data set; 4, performing fine tuning on the pre-trained VGG16 model by using LEMHI to obtain a LEMHI-CNN model; 5, inputting the video frame into a pre-trained VGG16 model to extract spatial features; 6, stacking, fragmenting and pooling the spatial features, and training an LSTM neural network model to obtain a CNN-LSTM model; 7, performing weighted fusion on the identification result of the LEMHI-CNN model and the CNN-LSTM model to obtain a final identification result. According to the invention, the video emotion recognition rate can be obviously improved.

Description

technical field [0001] The invention relates to a convolutional neural network, a cyclic neural network and classification discrimination, and belongs to the field of pattern recognition, in particular to a video emotion recognition method based on a dual-stream neural network. Background technique [0002] Traditional human-computer interaction, mainly through keyboards, mice, screens, etc., only pursues convenience and accuracy, and cannot understand and adapt to people's emotions and moods. Without this ability to understand and express emotions, it is difficult for a computer to have human-like intelligence. Emotion recognition is to endow computers with the ability to observe, understand and generate various emotional characteristics similar to humans, and finally enable computers to communicate and interact in a natural, friendly, and vivid way like humans. [0003] Research on video emotion recognition at home and abroad is generally divided into three steps: [000...

AI Technical Summary

Problems solved by technology

[0007] However, studies have shown that for face-related problems such as facial expression recognition and face detection, using the original image as input does not make good use of the motion information in the video, making the network's classification ability not high

And because the amount of data in the expression video dataset is small, it is easy to make the network overfitting

Images