Micro-expression recognition method based on adaptive motion amplification and convolutional neural network

Abstract

The invention discloses a micro-expression recognition method based on adaptive motion amplification and a convolutional neural network. The method comprises the following steps: 1, converting a certain sample of a micro-expression video into an image sequence, and carrying out the face cutting and alignment; 2, reading an initial frame of the image sequence, and calculating by using a vertex frame positioning algorithm to obtain a vertex frame picture; 3, determining a proper magnification factor by adopting a self-adaptive motion magnification method, and performing motion magnification on the vertex frame according to the determined magnification factor so as to enhance the characteristics of the micro-expression; 4, obtaining optical flow characteristics of the micro-expression video according to the start frame and the amplified vertex frame, and obtaining a horizontal optical flow, a vertical optical flow and optical strain; 5, establishing a convolutional neural network model for micro expression recognition, and performing transfer learning from macro expression to micro expression by using the model; and 6, inputting the optical flow features into the model after transfer learning, outputting time-space features, and training the model to realize micro-expression recognition.

Description

technical field [0001] The invention relates to a micro-expression recognition method based on adaptive motion amplification and a convolutional neural network, and belongs to the technical field of image processing. Background technique [0002] Micro Expression (ME) of human face is a facial expression with short duration, small range of muscle movement and uncontrollable. It usually occurs when human beings try to hide their true inner emotions, and it is not deceptive. The generation of micro-expressions is unconscious, but they can often effectively express a person's true emotions. Therefore, micro-expressions, as a clue to identify lies, have been extensively studied and applied in the fields of psychology, criminal investigation and security. Compared with macro-expressions, micro-expressions tend to occur quickly and last for a short time. Relevant studies have shown that their duration is usually 1 / 25 to 1 / 5 second; There will be 1 to 4 small areas in the face mot...

AI Technical Summary

Problems solved by technology

However, the above method uses the motion amplification technique to indiscriminately amplify all samples, although the micro-expression features after amplification are more obvious, it ignores the situation that the intensity of each micro-expression sample is different

Moreover, most of the existing micro-expression recognition methods based on video motion amplification technology use Eulerian Video Magnification (EVM) technology, which requires manual design of parameters, and the process is complicated. Prone to noise or excessive blur

Images