CN110826444A

CN110826444A - Facial expression recognition method and system based on Gabor filter

Info

Publication number: CN110826444A
Application number: CN201911032232.4A
Authority: CN
Inventors: 郑龙海
Original assignee: Beijing Yingpu Technology Co Ltd
Current assignee: Beijing Yingpu Technology Co Ltd
Priority date: 2019-10-28
Filing date: 2019-10-28
Publication date: 2020-02-21

Abstract

The application provides a facial expression recognition method and a system based on a Gabor filter, wherein in the method provided by the application, a user image is collected through video collection equipment, and the user image is preprocessed to obtain a face area in the user image; extracting a face feature image for face feature recognition by using a Gabor filter; and then inputting the facial feature image into a preset artificial neural network model, carrying out expression classification on the facial feature image, and identifying and outputting a facial expression identification result of the user image. Based on the method and the system for recognizing the facial expressions based on the Gabor filter, the method and the system for recognizing the facial expressions are provided, the facial expression classification is calculated by utilizing statistics and spatial characteristics such as entropy, moment, energy, mean value, variance and standard deviation, a neural network and the like, the robustness and the robustness of an algorithm are improved, and the execution time of the algorithm is compressed.

Description

Facial expression recognition method and system based on Gabor filter

Technical Field

The application relates to the technical field of face recognition, in particular to a facial expression recognition method and system based on a Gabor filter.

Background

In face Recognition (Facial Recognition), a Facial image of a user is acquired through video acquisition equipment, the position, the face shape and the angle of the Facial features of the user are calculated and analyzed by using a core algorithm, and then the Facial features are compared with an existing template in a database of the user, and then the real identity of the user is judged. The facial expression recognition is to separate a specific expression state from a facial image of a user so as to determine the psychological emotion of the user, and the main application fields of the technology comprise the fields of human-computer interaction, intelligent control, safety, medical treatment, communication and the like.

At present, a deep learning network is widely applied to a face recognition technology, and a model of the deep learning network is hierarchical and has large parameter capacity, so that data characteristics can be better displayed. The CNN has invariance of rotation, translation and scaling of spatial positions in image processing in deep learning, and can avoid the influence of face translation and deformation of other forms on recognition in face recognition, but the traditional classification face detection CNN has an overfitting problem and is low in recognition efficiency.

Disclosure of Invention

It is an object of the present application to overcome the above problems or to at least partially solve or mitigate the above problems.

According to one aspect of the application, a facial expression recognition method based on a Gabor filter is provided, and comprises the following steps:

acquiring a user image through video acquisition equipment, and preprocessing the user image to acquire a face area in the user image;

extracting a face feature image for face feature recognition based on the face region by using a Gabor filter;

and inputting the facial feature image into a preset artificial neural network model, carrying out expression classification on the facial feature image by the neural network model, and identifying and outputting a facial expression identification result of the user image.

Optionally, the acquiring, by a video capture device, a user image, and preprocessing the user image to obtain a face region in the user image includes:

collecting a user image through video collection equipment;

and preprocessing the user image by using a Viola-Jones face detection algorithm to obtain a face area in the user image.

Optionally, the preprocessing the user image by using a Viola-Jones face detection algorithm to obtain a face region in the user image includes:

establishing an integral image of the user image, and acquiring a Harr-like characteristic set of the user image based on the integral image;

training the Harr-like feature set through an Adaboost algorithm, and extracting facial features in the user image;

and screening the extracted facial features by using a cascade classifier to obtain a facial region in the user image.

Optionally, the extracting, by using a Gabor filter, a face feature image for face feature recognition based on the face region includes:

extracting Gabor characteristics used for face characteristic recognition in the face region through a Gabor filter according to the face region obtained after preprocessing and different directions and frequencies, and generating a face characteristic image of the user;

wherein, the Gabor filter is composed of a multi-resolution structure with a plurality of frequencies and a plurality of directions.

Optionally, the artificial neural network model is composed of a plurality of layers of feedforward networks;

the inputting of the facial feature image into a preset artificial neural network model, the expression classification of the facial feature image by the neural network model, and the recognition and output of the facial expression recognition result of the user image comprise:

and inputting the facial feature image into the preset artificial neural network model, carrying out pattern recognition based on the facial image features by a multilayer feed-forward network back propagation algorithm of the artificial neural network model, classifying and judging the expression type of the user in the user image, and outputting a facial expression recognition result.

According to another aspect of the present application, there is provided a Gabor filter-based facial expression recognition system, including:

the system comprises a face region acquisition module, a face region acquisition module and a face region processing module, wherein the face region acquisition module is configured to acquire a user image through a video acquisition device and preprocess the user image to acquire a face region in the user image;

a face feature image extraction module configured to extract a face feature image for face feature recognition based on the face region using a Gabor filter;

and the facial expression recognition module is configured to input the facial feature image into a preset artificial neural network model, perform expression classification on the facial feature image by using the neural network model, and recognize and output a facial expression recognition result of the user image.

Optionally, the facial region acquisition module is further configured to:

collecting a user image through video collection equipment;

Optionally, the facial region acquisition module is further configured to:

Optionally, the facial feature image extraction module is further configured to:

the facial expression recognition module is further configured to:

The application provides a facial expression recognition method and a system based on a Gabor filter, wherein in the method provided by the application, a user image is collected through video collection equipment, and the user image is preprocessed to obtain a face area in the user image; extracting a face feature image for face feature recognition by using a Gabor filter; and then inputting the facial feature image into a preset artificial neural network model, carrying out expression classification on the facial feature image, and identifying and outputting a facial expression identification result of the user image. According to the method and the system for recognizing the facial expressions based on the Gabor filter, statistics and spatial features such as entropy, moment, energy, mean value, variance and standard deviation, a neural network and the like are used for calculating facial expression classification, so that the robustness of the algorithm is improved, the execution time of the algorithm is reduced, the facial expression recognition efficiency is further improved, and human-computer interaction is realized more intelligently.

The above and other objects, advantages and features of the present application will become more apparent to those skilled in the art from the following detailed description of specific embodiments thereof, taken in conjunction with the accompanying drawings.

Drawings

Some specific embodiments of the present application will be described in detail hereinafter by way of illustration and not limitation with reference to the accompanying drawings. The same reference numbers in the drawings identify the same or similar elements or components. Those skilled in the art will appreciate that the drawings are not necessarily drawn to scale. In the drawings:

fig. 1 is a schematic flow chart of a Gabor filter-based facial expression recognition method according to an embodiment of the present application;

FIG. 2 is a schematic flow chart of the Viola-Jones face detection algorithm according to an embodiment of the present application;

FIG. 3 is a schematic structural diagram of a Gabor filter-based facial expression recognition system according to an embodiment of the present application;

FIG. 4 is a schematic diagram of a computing device according to an embodiment of the present application;

FIG. 5 is a schematic diagram of a computer-readable storage medium according to an embodiment of the present social situation.

Detailed Description

Fig. 1 is a schematic flow chart of a facial expression recognition method based on a Gabor filter according to an embodiment of the present application. As can be seen from fig. 1, the method for recognizing facial expressions based on a Gabor filter according to the embodiment of the present application may include:

step S101: acquiring a user image through video acquisition equipment, and preprocessing the user image to acquire a face area in the user image;

step S102: extracting a face feature image for face feature recognition based on the face region by using a Gabor filter;

step S103: and inputting the facial feature image into a preset artificial neural network model, carrying out expression classification on the facial feature image by the neural network model, and identifying and outputting a facial expression identification result of the user image.

The embodiment of the application provides a facial expression recognition method based on a Gabor filter, wherein in the method provided by the embodiment of the application, a user image is collected through video collection equipment, and the user image is preprocessed to obtain a face area in the user image; extracting a face feature image for face feature recognition by using a Gabor filter; and then inputting the facial feature image into a preset artificial neural network model, carrying out expression classification on the facial feature image, and identifying and outputting a facial expression identification result of the user image. According to the method and the system for recognizing the facial expression based on the Gabor filter, the Gabor filter is used for extracting facial features, and the multilayer artificial neural network is used for carrying out pattern recognition on the facial expression, so that the robustness of the algorithm is improved, the execution time of the algorithm is compressed, the facial expression recognition efficiency is further improved, and the man-machine interaction is realized more intelligently.

In image processing, the Gabor function is a linear filter for edge extraction. The frequency and directional representation of the Gabor filter is similar to the human visual system. The Gabor filter was found to be well suited for texture expression and separation. In the spatial domain, a two-dimensional Gabor filter is a gaussian kernel modulated by a sinusoidal plane wave.

Referring to step S101, a user image is captured by a video capture device to be preprocessed, so as to obtain a face region in the user image. The video capture device may include a video camera, a still camera, or the like. Optionally, the step may further comprise: collecting a user image through video collection equipment; and preprocessing the user image by using a Viola-Jones face detection algorithm to obtain a face area in the user image.

Preprocessing is the first step of face detection. In most cases, when a photograph is taken, the face area is not the only area to be taken, and the photograph may include a background other than a human face. Therefore, a human face is required as a region of interest for feature extraction. The region of interest (ROI) is new data that is put into a block of original image data before releasing a selected block of region that needs to be used. The ROI must be selected within the original image. The region of interest in the embodiment of the present application refers to a face region of a user. That is, in the present embodiment, the region of interest is only a face image on the entire image, such as reducing the processing time and increasing the reliability by removing the background as unnecessary information, for example, the background and hair of the image in the JAFFE database do not provide information in recognizing expressions (the JAFFE database is one of the most widely used databases in the expression recognition field, and contains 7 expressions in total).

As introduced above, the user image may be pre-processed using the Viola-Jones face detection algorithm. The Viola-Jones algorithm is a method for performing face detection based on Haar feature values of a face. For feature extraction, Haar-like features are rectangular types obtained by integrating images. The Haar feature used by Viola-Jones is related only to the sum of the pixel values within a rectangular area, calculated by subtracting the sum of the white rectangles from the sum of the black rectangles. The Haar-like features reflect the gray level change of the image, and the human face features are quantized to distinguish human faces from non-human faces. As shown in fig. 2, the face detection method may include the following steps:

step S201: establishing an integral image of a user image, and acquiring a Harr-like characteristic set of the user image based on the integral image;

step S202: training the Harr-like feature set through an Adaboost algorithm, and extracting facial features in the user image;

step S203: and screening the extracted facial features by using a cascade classifier to obtain a facial region in the user image.

In the face detection algorithm, an integral image of the user image is created based on the input image, as described in step S201. Integral images are a method of rapidly computing the sum of rectangular regions in an image, which refers to the sum of all pixels above and to the left of the associated pixel. The main advantage of this algorithm is that once the integral image is first calculated, the sum of rectangular regions of arbitrary size in the image can be calculated in a constant time. Therefore, the calculation amount is greatly reduced and the calculation speed is improved during image blurring, edge extraction and object detection.

Step S202 is performed next to extract features from the Harr-like feature set, i.e. the sub-window of the user image. And training and calculating the Harr-like feature set by using an Adaboost algorithm, and extracting facial features in the user image. The basic size of the sub-window of the user image is 24 x 24 pixels, each element type is scaled and shifted in all possible combinations. In a 24 x 24 pixel sub-window, there are approximately 160,000 possible features to compute. Therefore, the embodiments of the present application use Adaboost (i.e., machine learning algorithm), which helps to find the best function among all of these 160,000 functions, also referred to as weak classifiers. The Adaboost algorithm is an iterative algorithm, and the core idea is to train different classifiers (weak classifiers) aiming at the same training set, and then to assemble the weak classifiers to form a stronger final classifier (strong classifier).

Haar-like features (Haar-like features) are a type of digital image feature used for object recognition that uses adjacent rectangles at a given location in a detection window to compute the sum of pixels for each rectangle and take the difference. These differences are then used to classify the sub-regions of the image. Harr features were first used for face representation. In the detection phase, a detection window of the same size as the target object is slid over the input image, and a Harr feature is calculated for each region of the image. These differences are compared to a pre-calculated threshold to distinguish between targets and non-targets. Since such a Harr feature is a weak classifier, a large cluster of such features is required to arrive at a reliable decision, and these features are combined into a cascade of classifiers, eventually forming a strong classifier. The main advantage of the Harr feature is that it is extremely fast to compute, and with a structure called an integral graph, Harr features of any size can be computed in a constant time. Harr characteristics are classified into 3 categories: and combining the edge characteristic, the linear characteristic, the central characteristic and the diagonal characteristic into a characteristic template. The characteristic template is provided with two rectangles of white and black, and the characteristic value of the template is defined as a white rectangle pixel sum minus a black rectangle characteristic pixel sum. The Harr characteristic reflects the gray scale change of the image. However, the rectangular feature is sensitive to some simple graphic structures such as edges and line segments, so that only structures with specific trends (horizontal, vertical and diagonal) can be described.

After the facial features in the user image are extracted, the last step S203 in the Viola-Jones face detection algorithm is executed, and the extracted facial features are screened by using a cascade classifier to obtain the facial region in the user image. It should be noted that. In this embodiment, all english words appearing in the Gabor filter, the Adaboost algorithm, the Viola-Jones face detection algorithm, and the like belong to proper nouns, and there is no proper chinese paraphrase, and those skilled in the art can obtain the chinese meaning indicated by the above english words based on the corresponding english nouns.

The "weak" classifiers are linearly combined into a "strong" classifier using a cascade of classifiers. The basic principle of the Viola-Jones face detection algorithm is to scan the same image multiple times, each time with a new size. If the image contains one or more faces, it is clear that the sub-window evaluated should still be non-faces. Therefore, the algorithm should focus on quickly discarding non-faces and spending more time on possible face regions.

After the user image is preprocessed, referring to step S102, a Gabor filter is used to extract a face feature image for face feature recognition. The Gabor filter is a linear filter for edge extraction, has frequency and direction expression similar to those of the human visual system, can provide good direction selection and scale selection characteristics, and is insensitive to illumination change.

In an alternative embodiment of the present application, extracting the feature image by using a Gabor filter may include: extracting Gabor characteristics used for face characteristic identification in the face area by using different directions and frequencies through a Gabor filter based on the face area obtained after preprocessing, and generating a face characteristic image of the user; wherein, the Gabor filter is composed of a multi-resolution structure with a plurality of frequencies and a plurality of directions.

The Gabor filter can extract both the temporal (spatial) and frequency domains. It is used in many applications such as object detection, fingerprint recognition, image analysis and compression, edge detection, document analysis, character recognition, texture segmentation and classification, face recognition and iris detection, etc.

Gabor filters have proven successful in FER (Facial expression recognition). The use of the Gabor filter by the multi-resolution structure means a filter composed of multiple frequencies and multiple directions. The multiresolution structure of the Gabor filter correlates the Gabor filter with the wavelet. Used in the examples of the present application are 2D-Gabor (two-dimensional Gabor) filters, which are defined as follows:

wherein:

(x, y) represents a pixel position in the spatial domain;

θ represents the direction of the Gabor filter;

ω denotes the radial center frequency, and σ is the standard deviation of the Gaussian function along the x-axis and the y-axis;

Ψ denotes a phase offset;

x 'and y' may represent the associated parametric equations, respectively.

The main parameters of a Gabor filter are its frequency and direction. For feature extraction of a facial image, the image is convolved with a Gabor filter as follows:

wherein:

(x, y) represents a pixel position in the spatial domain;

θ represents the direction of the Gabor filter;

i (x, y) represents an operator function equation with respect to x, y.

As mentioned above, Gabor filters extract Gabor features by using different directions and frequencies. In general, a Gabor filter bank extracts Gabor features for facial feature recognition using five frequencies and eight directions.

Selection of any discrete rotation theta_iAnd the calculated directions must be evenly distributed as follows:

wherein O represents the total number of directions.

To reduce complexity and computation to half, we use the angular responses [ pi, 2 pi ] because they are complex conjugates of the responses of [0, pi ] with real-valued inputs. The frequency calculation formula is as follows:

wherein: s represents a scale.

Thus, a total of 40Gabor features are used to decompose and obtain multi-frequency (scale) and multi-directional Gabor features. Embodiments of the present application prefer to use true Gabor amplitude information to characterize the smoothness and stability of human facial expressions. Therefore, a 40Gabor amplitude image (GMP) is characteristic of Gabor.

And finally, executing the step S103, inputting the facial feature image into a preset artificial neural network model to perform expression classification on the facial feature image, and identifying and outputting a facial expression identification result of the user image.

In an alternative embodiment of the present application, the artificial neural network model is constructed from a multi-layer feedforward network. Through artificial neural network model discernment facial expression, still include: inputting the facial feature image into a preset artificial neural network model, carrying out pattern recognition based on facial image features by a multilayer feedforward network back propagation algorithm of the artificial neural network model, classifying and judging the expression type of a user in the user image, and outputting a facial expression recognition result.

Artificial Neural Networks (ANN) are an example of information processing that is inspirational to process information from biological nervous systems, such as the brain. A biological neural network consisting of a number of nerve cells (neurons) interconnected by synapses. Neurons have an output called dendrites. Axons transmit their nerve impulses through neuromuscular junctions with the help of synapses (neurotransmitters) located at the axon terminals, which is the simplest model of biological neural networks.

The artificial neural network is an artificial neural network, which is formed by using a single perceptron as a neural network node and then using the node to form a hierarchical network structure. When the network has more than or equal to 3 layers (input layer + middle layer (more than or equal to 1) + output layer), the network is called a multilayer artificial neural network.

The artificial neural network is a multilayer artificial neural network, which is an extension of a single-layer network and includes an input layer, an output layer and one or more intermediate layers. These intermediate layers are called "hidden" layers because their activities are not accessible from outside the network as the input and output layers. Without specific rules and theorems, the number of neurons in the hidden layer can be difficult to determine, and successful approaches have been implemented in the fields of approximation, classification, prediction, and pattern recognition.

The artificial neural network model can be divided into two types of models according to the topological structure, namely a feed-forward (feed-forward) model and a feed-back (feed-back) model, the multilayer artificial neural network preset in the implementation of the application uses a multilayer feed-forward network, the feed-forward neural network is the simplest neural network, each neuron is arranged in a layered mode, each neuron is only connected with the neuron of the previous layer, namely, the signal is allowed to propagate from the input to the output from only one direction, the output of the previous layer is received and output to the next layer, and no feedback (circulation) exists between the layers. Feed-forward ANNs tend to be direct networks that associate inputs with outputs. They are widely used for pattern recognition (pattern recognition is the most important application of artificial neural networks), this type of organization also being referred to as bottom-up or top-down.

The feedforward ANN of the present network employs a back propagation algorithm. The back propagation algorithm is to train the feedforward neural network better and faster to obtain the weight parameters and bias parameters of each layer of the neural network, and it adopts the gradient descent method to try to minimize the square error between the network output value and the target value.

The back propagation algorithm mentioned above is required to be trained and tested in advance, and involves a predetermined set of data as basic data for training and testing. During the training process, the network is excited by a set of predefined data, called the training set. The training process of the training set typically includes the following four steps:

s1: assembling training data;

s2: creating a network object;

s3: training a network;

s4: simulating the response of the network to the new input;

after training the data, to achieve classification of the predetermined expressions, user definitions targeting seven expressions are needed, including: [1000000] for anger, [0100000] for aversion, [0010000] for fear, [0001000] for happiness, [0000100] for neutrality, [0000010] for sadness and [0000001] for surprise.

That is, the input facial feature image is recognized through a preset multilayer feedforward ANN, the user image is judged according to the targets of seven expressions, one expression is matched, and the user expression recognition result is output.

Based on the same inventive concept, as shown in fig. 3, the embodiment of the present application further provides a facial expression recognition system based on a Gabor filter, which includes

A face region acquisition module 310 configured to acquire a user image through a video acquisition device, and pre-process the user image to acquire a face region in the user image;

a face feature image extraction module 320 configured to extract a face feature image for face feature recognition based on the face region using a Gabor filter;

and the facial expression recognition module 330 is configured to input the facial feature image into a preset artificial neural network model, perform expression classification on the facial feature image by using the neural network model, and recognize and output a facial expression recognition result of the user image.

In an optional embodiment of the present application, the facial region acquisition module 310 is further configured to:

collecting a user image through video collection equipment;

the user image is preprocessed using the Viola-Jones face detection algorithm to obtain the face region in the user image.

In an optional embodiment of the present application, the facial region obtaining module 310 may be further configured to:

In an optional embodiment of the present application, the facial feature image extraction module 320 is further configured to:

extracting Gabor characteristics used for face characteristic identification in the face area by using different directions and frequencies through a Gabor filter based on the face area obtained after preprocessing, and generating a face characteristic image of the user; wherein, the Gabor filter is composed of a multi-resolution structure with a plurality of frequencies and a plurality of directions.

In an alternative embodiment of the present application, the artificial neural network model is composed of a multi-layer feedforward network;

the facial expression recognition module 330, which is further configured to:

and inputting the facial feature image into a preset artificial neural network model, carrying out pattern recognition based on facial image features by a multilayer feed-forward network back propagation algorithm of the artificial neural network model, classifying and judging the expression type of the user in the user image, and outputting a facial expression recognition result.

The application provides a facial expression recognition method and a system based on a Gabor filter, wherein in the method provided by the application, a user image is collected through video collection equipment, and the user image is preprocessed to obtain a face area in the user image; extracting a face feature image for face feature recognition by using a Gabor filter; and then inputting the facial feature image into a preset artificial neural network model, carrying out expression classification on the facial feature image, and identifying and outputting a facial expression identification result of the user image.

According to the method and the system for recognizing the facial expressions based on the Gabor filter, statistics and spatial features such as entropy, moment, energy, mean value, variance and standard deviation, a neural network and the like are used for calculating facial expression classification, so that the robustness of the algorithm is improved, the execution time of the algorithm is reduced, the facial expression recognition efficiency is further improved, and human-computer interaction is realized more intelligently.

Embodiments of the present application further provide a computing device, referring to fig. 4, comprising a memory 420, a processor 410, and a computer program stored in the memory 420 and executable by the processor 410, the computer program being stored in a space 430 for program code in the memory 420, the computer program, when executed by the processor 410, implementing steps 431 for performing any of the methods according to the present invention.

The embodiment of the application also provides a computer readable storage medium. Referring to fig. 5, the computer readable storage medium comprises a storage unit for program code provided with a program 431' for performing the steps of the method according to the invention, which program is executed by a processor.

The embodiment of the application also provides a computer program product containing instructions. Which, when run on a computer, causes the computer to carry out the steps of the method according to the invention.

In the above embodiments, the implementation may be wholly or partially realized by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loaded and executed by a computer, cause the computer to perform, in whole or in part, the procedures or functions described in accordance with the embodiments of the application. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable device. The computer instructions may be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another, for example, from one website site, computer, server, or data center to another website site, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device, such as a server, a data center, etc., that incorporates one or more of the available media. The usable medium may be a magnetic medium (e.g., floppy Disk, hard Disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., Solid State Disk (SSD)), among others.

Those of skill would further appreciate that the various illustrative components and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.

It will be understood by those skilled in the art that all or part of the steps in the method for implementing the above embodiments may be implemented by a program, and the program may be stored in a computer-readable storage medium, where the storage medium is a non-transitory medium, such as a random access memory, a read only memory, a flash memory, a hard disk, a solid state disk, a magnetic tape (magnetic tape), a floppy disk (floppy disk), an optical disk (optical disk), and any combination thereof.

The above description is only for the preferred embodiment of the present application, but the scope of the present application is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present application should be covered within the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A facial expression recognition method based on a Gabor filter comprises

2. The method of claim 1, wherein the capturing of the user image by the video capture device, the pre-processing of the user image to obtain the facial region in the user image, comprises:

collecting a user image through video collection equipment;

3. The method of claim 2, wherein said pre-processing said user image using Viola-Jones face detection algorithm to obtain a facial region in said user image comprises:

4. The method according to claim 1, wherein the extracting, by using a Gabor filter, a face feature image for face feature recognition based on the face region comprises:

5. The method of claim 1, wherein the artificial neural network model is comprised of a multi-layer feed-forward network;

6. A facial expression recognition system based on a Gabor filter comprises

7. The system of claim 6, wherein the facial region acquisition module is further configured to:

collecting a user image through video collection equipment;

8. The system of claim 7, wherein the facial region acquisition module is further configured to:

9. The system of claim 6, wherein the facial feature image extraction module is further configured to:

10. The system of claim 6, wherein the artificial neural network model is comprised of a multi-layer feed-forward network;

the facial expression recognition module is further configured to: