Facial expression recognition training system and facial expression recognition training method
A facial expression recognition training system includes a training module, feature database, a capturing module, a recognition module and an adjusting module. The training module trains a facial expression feature capturing model according to known face images. The feature database stores known facial expression features of the known face images. The capturing module continuously captures first face images, and the facial expression feature capturing model outputs facial expression features of the first face images according to the first face images. The recognition module compares the facial expression features and the known facial expression features, and fit the facial expression features to the first known facial expression features that is one kind of the known facial expression feature accordingly. The adjusting module adjusts the facial expression feature capturing model to reduce the differences between the facial expression features and the known facial expression features.
1. A facial expression recognition training system, characterized by comprising:
a training module configured to train a facial expression feature capturing model according to a plurality of known facial images;
a feature database configured to store a plurality of known facial expression features of the plurality of known facial images;
a capturing module configured to continuously capture a plurality of first facial images, wherein the facial expression feature capturing model outputs a plurality of facial expression features of the plurality of first facial images according to the plurality of first facial images;
a recognition module configured to compare the plurality of facial expression features with the plurality of known facial expression features, and make the plurality of facial expression features correspond to the plurality of known facial expression features accordingly;
an adjusting module configured to adjust the facial expression feature capturing model to reduce differences between the plurality of facial expression features and the plurality of known facial expression features corresponding to the plurality of facial expression features; and
an image calibration module configured to correct each of the plurality of first facial images to form a second facial image, and sharpen the second facial image, wherein the facial expression feature capturing model recognizes and outputs the plurality of facial expression feature to the recognition module according to the sharpened second facial image, and the recognition module compares the plurality of facial expression feature which is processed by the image calibration module with the plurality of known facial expression features.
2. The facial expression recognition training system of claim 1 , characterized in that the facial expression feature capturing model comprises a convolutional neural network or a neural network.
3. The facial expression recognition training system of claim 1 , characterized in that the image calibration module is configured to align facial features of each of the plurality of first facial images with each other.
4. The facial expression recognition training system of claim 1 , characterized in that the image calibration module is configured to shift each of the plurality of first facial images with respect to a reference plane such that a nosal tip feature point in each of the plurality of first facial images is aligned with a center point of the reference plane, and rotate each of the plurality of shifted first facial images such that a connection line between two eyes in the each of the plurality of shifted first facial images is parallel to a horizontal line of the reference plane, so as to form the second facial image.
5. A facial expression recognition training method, characterized by comprising:
training a facial expression feature capturing model according to a plurality of known facial images;
storing a plurality of known facial expression features of the plurality of known facial images;
capturing a plurality of first facial images;
outputting a plurality of facial expression features according to the plurality of first facial images;
comparing the plurality of facial expression features with the plurality of known facial expression features, and making the plurality of facial expression features correspond to the plurality of known facial expression features accordingly; and
adjusting the facial expression feature capturing model to reduce differences between the plurality of facial expression features and the plurality of known facial expression features corresponding to the plurality of facial expression features, wherein the step of outputting the facial expression features according to the first facial images comprises:
correcting each of the plurality of first facial images with respect to a reference plane to form a second facial image;
sharpening the second facial image; and
recognizing the plurality of facial expression features according to the sharpened second facial image.
6. The facial expression recognition training method of claim 5 , characterized in that the step of sharpening the second facial image comprises:
using Difference of Gaussian (DoG), Sobel operator or Laplace operator to sharpen the second facial image.
7. The facial expression recognition training method of claim 5 , characterized in that the facial expression feature capturing model comprises a convolutional neural network or a neural network.
8. The facial expression recognition training method of claim 5 , characterized in that the step of correcting the plurality of first facial images with respect to the reference plane to form the second facial image comprises:
shifting the plurality of first facial images with respect to the reference plane such that a nosal tip feature point in the plurality of first facial images is aligned with a center point of the reference plane; and
rotating the plurality of shifted first facial images such that a connection line between two eyes in the each of the plurality of shifted first facial images is parallel to a horizontal line of the reference plane, so as to form the second facial image.