Lie-telling detection method based on micro-expressions in interview

Abstract

The invention relates to a lie-telling detection method based on micro-expressions in an interview, which comprises the following steps: firstly, training five expressions, namely eyebrow wrinkling, eyebrow lifting, mouth closing, pouting and head tilting, by a model, and labeling each type of expression data; secondly, inputting an image of the facial micro-expression into a pre-trained SSD network taking VGG16 as a backbone, enabling the image to pass through a convolutional neural network to extract features, and generating a feature map; performing convolution operation on each feature mapto evaluate a default bounding box, and predicting an offset and a classification probability for each bounding box; combining bounding boxes obtained by different feature maps, executing a non-maximum suppression method to filter a part of overlapped or incorrect borders, and generating a final bounding box set; and finally, classifying detection results by using a classifier. According to the method, high-level and low-level visual features are used at the same time, and compared with human beings, the method is obviously better in cheating prediction; and compared with naked eye judgment of human beings, the speed is higher, and the technical accuracy is higher.

Description

technical field [0001] The invention relates to a method for detecting lying based on micro-expressions in interviews, and belongs to the technical field of image information processing. Background technique [0002] Lying and deception often occur during interviews, and how to effectively identify lies is very important. Although lying is a common phenomenon, the micro-expressions of people when lying are different from those in normal conversation. Some small changes in expressions and subtle muscle twitches on the face can easily reveal the true thoughts unconsciously. Micro-expression experts can capture these micro-expressions, and then judge whether there is lying and deceit in the interview process. [0003] However, because the interview process is susceptible to interference from various factors, the accuracy of relying on manual detection of lying is often greatly reduced, and traditional robot face recognition is prone to interference from other facial features b...

AI Technical Summary

Problems solved by technology

Although lying is a common phenomenon, people's micro-expressions when lying are different from those in normal conversations. Some small changes in expressions and subtle muscle twitches on the face can easily reveal the true thoughts unconsciously. of micro-expression experts can capture these micro-expressions, and then judge whether there is lying and deceit in the interview process

[0003] However, because the interview process is easily interfered by various factors, the accuracy of relying on manual detection of lying is often greatly reduced, and traditional robot face recognition is prone to interference from other facial features because it does not specifically optimize micro-expressions , cannot be directly applied to micro-expression judgment

Images