An intelligent gender identification method based on facial images using compound modules

Abstract

The present invention relates to an automated system and method for gender classification using facial recognition technology. The system utilizes a two-module approach, comprising an Automated Face Identifier (AFI) and an Automated Gender Identifier (AGI). The AFI module detects a human face from an input image, identifies key facial landmarks (such as the eyes, nose, and mouth), and extracts facial features using frequency domain analysis techniques, including Fast Fourier Transform (FFT) and Discrete Cosine Transform (DCT). The AGI module classifies the gender of the individual based on the extracted features using machine learning algorithms, such as Support Vector Machine (SVM), Decision Tree (DT), and Random Forest (RF). The system can process both static images and video frames, offering a highly accurate gender classification with up to 97.12% accuracy using the Random Forest algorithm. This technology has broad applications in fields such as security, surveillance, and human-computer interaction, providing an efficient and scalable solution for real-time gender prediction.

Specification

Description:
Field of the Invention:
[001] The field of the invention pertains to automated gender detection and classification using facial recognition technology. More specifically, it involves the development of advanced systems and methods that leverage artificial intelligence, machine learning, and deep learning techniques to accurately predict a person's gender based on facial features. This invention is primarily focused on the use of frontal facial images, incorporating both temporal and spectral domain analysis for enhanced prediction accuracy. The system is designed to detect faces and extract relevant facial attributes, which are then used to classify the gender of individuals. The technology has wide applications in various industries such as security, surveillance, human-computer interaction, and biometric identification, offering a reliable and efficient tool for gender recognition in both static images and video files.

Background of the invention and related prior art:
[002] The background of the invention lies in the growing need for accurate and efficient gender detection systems within the fields of facial recognition and biometric analysis. Over the years, advancements in computer vision, machine learning, and artificial intelligence have enabled the development of systems capable of recognizing and classifying various human attributes from images, including age, ethnicity, and gender. Gender detection, in particular, has become an important aspect of identity verification, human-computer interaction, and security applications. Traditional methods relied heavily on manual classification or basic algorithms, often yielding inconsistent or inaccurate results. With the rise of deep learning and advanced facial feature extraction techniques, it is now possible to achieve high levels of accuracy in gender prediction. This invention seeks to address these challenges by combining cutting-edge machine learning models with temporal and spectral analysis to create a robust system capable of identifying gender from both images and video files, enhancing the capabilities and reliability of automated gender classification systems.
[003] A patent document US8027521B1 discloses a method and system to provide a face-based automatic gender recognition system that utilizes localized facial features and hairstyles of humans. Given a human face detected from a face detector, it is accurately localized to facilitate the facial/hair feature detection and localization. Facial features are more finely localized using the geometrically distributed learning machines. Then the position, size, and appearance information of the facial features are extracted. The facial feature localization essentially decouples geometric and appearance information about facial features, so that a more explicit comparison can be made at the recognition stage. The hairstyle features that possess useful gender information are also extracted based on the hair region segmented, using the color discriminant analysis and the estimated geometry of the face. The gender-sensitive feature vector, made up from the extracted facial and hairstyle features, is fed to the gender recognition machines that have been trained using the same kind of gender-sensitive feature vectors of gallery images.
[004] Another patent document US9767349B1 discloses a method for determining an emotional state of a subject taking an assessment. The method includes eliciting predicted facial expressions from a subject administered questions each intended to elicit a certain facial expression that conveys a baseline characteristic of the subject; receiving a video sequence capturing the subject answering the questions; determining an observable physical behavior experienced by the subject across a series of frames corresponding to the sample question; associating the observed behavior with the emotional state that corresponds with the facial expression; and training a classifier using the associations. The method includes receiving a second video sequence capturing the subject during an assessment and applying features extracted from the second image data to the classifier for determining the emotional state of the subject in response to an assessment item administered during the assessment.
[005] A document US7912246B1 discloses a system and method for performing age classification or age estimation based on the facial images of people, using multi-category decomposition architecture of classifiers. In the multi-category decomposition architecture, which is a hybrid multi-classifier architecture specialized to age classification, the task of learning the concept of age against significant within-class variations, is handled by decomposing the set of facial images into auxiliary demographics classes, and the age classification is performed by an array of classifiers where each classifier, called an auxiliary class machine, is specialized to the given auxiliary class. The facial image data is annotated to assign the gender and ethnicity labels as well as the age labels. Each auxiliary class machine is trained to output both the given auxiliary class membership likelihood and the age group likelihoods. Faces are detected from the input image and individually tracked. Age sensitive feature vectors are extracted from the tracked faces and are fed to all of the auxiliary class machines to compute the desired likelihood outputs. The outputs from all of the auxiliary class machines are combined in a manner to make a final decision on the age of the given face.
[006] Another document US9317785B1 discloses a system and method for performing ethnicity classification based on the facial images of people, using multi-category decomposition architecture of classifiers, which include a set of predefined auxiliary classifiers that are specialized to auxiliary features of the facial images. In the multi-category decomposition architecture, which is a hybrid multi-classifier architecture specialized to ethnicity classification, the task of learning the concept of ethnicity against significant within-class variations, is handled by decomposing the set of facial images into auxiliary demographics classes; the ethnicity classification is performed by an array of classifiers where each classifier, called an auxiliary class machine, is specialized to the given auxiliary class. The facial image data is annotated to assign the age and gender labels as well as the ethnicity labels. Each auxiliary class machine is trained to output both the given auxiliary class membership likelihood and the ethnicity likelihoods. Faces are detected from the input image, individually tracked, and fed to all the auxiliary class machines to compute the desired auxiliary class membership and ethnicity likelihood outputs. The outputs from all the auxiliary class machines are combined in a manner to make a final decision on the ethnicity of the given face.
[007] A patent document TR2021019127A2 related to the intelligent system that detects the suspects with the hybrid model of gait analysis and face recognition, which allows to predict the suspects from the video images of the ones whose faces are not visible in the video, from the gait of the suspects whose gait is visible only in a small frame and whose gait is not visible. Height, weight, eye color, hair color, hair type, skin color, beard type, mustache type, ethnicity, race, etc. recording soft biometric features, analyzing the faces and gaits of the people in the video via the desktop/mobile/web-based system developed and training the system and giving a new video recording that has not been given to the system before and the classification of the images of the unknown person in the video given to the system includes the process steps.
[008] None of these above patents, however alone or in combination, disclose the present invention. The invention consists of certain novel features and a combination of parts hereinafter fully described, illustrated in the accompanying drawings, and particularly pointed out in the appended claims, it being understood that various changes in the details may be made without departing from the spirit, or sacrificing any of the advantages of the present invention.

Summary of the invention:
[009] The invention provides an advanced system for automated gender detection and classification using facial recognition technology. It combines machine learning, deep learning, and temporal and spectral domain analysis to accurately predict a person's gender from frontal facial images. The system consists of two primary modules: the Automated Face Identifier (AFI), which detects faces and extracts facial features, and the Automated Gender Identifier (AGI), which classifies gender based on these extracted features. The invention utilizes a hybrid approach to enhance prediction accuracy, achieving a high classification rate of 97.12% with the use of the random forest algorithm. This system can process both static images and video files, making it applicable for a variety of practical uses such as security monitoring, surveillance, and human-computer interaction. The technology is designed to be both user-friendly and environmentally adaptable, offering an efficient solution for gender classification in real-time applications.

Detailed description of the invention with accompanying drawings:
[010] For the purpose of facilitating an understanding of the invention, there is illustrated in the accompanying drawing a preferred embodiment thereof, from an inspection of which, when considered in connection with the following description, the invention, its preparation, and many of its advantages should be readily understood and appreciated.
[011] The principal object of the invention is to develop an intelligent gender identification method based on facial images using compound modules. In this investigation, an input database including huge frontal facial photographs of male and female celebrities was utilized. This experimental dataset, CelebA, is freely accessible to the public. CelebFaces Attributes Collection (CelebA) is a substantial facial characteristics collection containing over 200K celebrity photos, each one with 40 attribute annotations. The photos in this collection span a wide range of position variants as well as complex backgrounds. CelebA offers a wide range of annotations, as well as a big number of them.
Method details
The most significant aspect of biometric identification is the face, which is a crucial component in distinguishing between male and female. To carry out this investigation two modules were implemented.
1) Automated Face Identifier (AFI),
2) Automated Gender Identifier (AGI).
Automated Face Identifier (AFI)
The Automated Face Identifier module consists of three major steps,
1) Stage 1: Face and Facial Parts Identification,
2) Stage 2: The important features extraction procedure, and
3) Stage 3: Construction of feature vectors.

Stage 1: Face and Facial Parts Identification
The identification of the face and facial parts is the first and most important step in this investigation. Since the frontal face of a person contains the majority of the significant traits, frontal face photos are favoured in this study. The initial step is to identify a human face from an input image and to accomplish this, the logic of the Viola-Jones algorithm is utilized.
This algorithm offers many benefits:
1. This method can distinguish between a face and a non-face from an unknown image.
2. The accuracy rate is significantly greater.
3. It produces findings with a larger proportion of true positives and a lower proportion of false positives.
4. It has numerous uses in real-time systems.
Following the face detection, the next stage is to detect the facial parts from the face region. The salient facial parts of the frontal face region include the eyes, nose, mouth, and two eyebrow regions. As a solution, the 68 facial landmark detection procedure was used to recognize these locations more precisely.
Figure 1. experimental findings and the methodology according to the embodiment of the present invention.
Stage 2: The Important Features Extraction Procedure
The frequency domain and spatial domain concepts are used at this stage to evaluate the features from the input data. The fast wavelet transform and Discrete Cosine Transformation evaluation are the finest options for this application.
Fast Fourier analysis (FFT) is an efficient method for determining a discrete sequence from an input. This algorithm converts a signal from its spatial domain to its frequency domain and vice versa. For N points, it reduces the number of assessments from to, where is the base-2 method.
Consider a_0to a_n to be complex numbers. As a result, FFT is defined as,
∑_(n=0)^(N-1)▒〖a_n e^(-2πink/N)= 〗 ∑_(n=0)^(N/2-1)▒〖a_2n e^(-2πi(2n)k/N)+〗 ∑_(n=0)^(N/2-1)▒〖a_(2n+1) e^(-2πi(2n+1)k/N) =〗 ∑_(n=0)^(N/2-1)▒〖a_n evene^(-2πink/(N/2))+ 〗 e^(-2πink/N) ∑_(n=0)^(N/2-1)▒〖a_n odde^(-2πink/(N/2)) 〗
Where, k = 0,…,N-1 and e^(-2πink/N) is a N^th primitive root of 1.
The benefit of employing FFT is that it aids in the conversion of time domain to frequency domain, making calculations faster. The second and most important advantage is that it can convert discrete data into a continuous data type that is accessible at wide bandwidth.
A discrete cosine transforms (DCT) represents a finite sequence of data points as a combination of cosine functions vibrating at various frequencies. It is used to compress images in most digital media, including digital images. This approach was utilized in this experiment to transform the pixel values of the input image from the spatial domain to the frequency domain. The equation for DCT evaluation is given below:
F(i,j)= 1/4 C_i C_j ∑_(x=0)^7▒∑_(y=0)^7▒〖f(i,j) cos⁡〖(2x+1)jπ/16〗 cos⁡〖(2y+1)iπ/16〗 〗
The benefits of employing the discrete cosine transform are as follows:
1) it is a quick transform,
2) it has great compression for closely correlated data, and
3) It fixes basic images and strikes a fine balance between information compressing capability and computational complexity.
Stage 3: Construction of Feature Vector
Two wavelet processing approaches were utilized in this investigation to extract features from identified regions and construct feature matrices for the following phase. It is the second principal technique used to carry out this investigation.
The feature matrices for the input images are generated by these two wavelet techniques. As a result, the Euclidean distance measurement approach is used in the following step to construct the feature vectors for the prediction model. This method calculates the distance and length between two locations.
If a region of interest has N number of points and (x1,y1) and (x2,y2) are the two points of it. As a result, the Euclidean distance is computed as follows:
Euclidean Distance,ED= √((〖x_2〗^2-〖x_1〗^2 )-(〖y_2〗^2-〖y_1〗^2 ) )
Automated Gender Identifier (AGI)
To identify the gender, five machine learning algorithms are used in this procedure. The models include Support Vector Machine (SVM), Decision Tree (DT), K-Nearest Neighbour (KNN), Naive Bayes (NB), and Random Forest (RF).
Support Vector Machine (SVM)
The Support Vector Machine (SVM) algorithm is a type of supervised machine learning algorithm. The SVM algorithm's goal is to find a hyperplane in a high dimensional space that clearly separates the input points. Whenever there is a substantial margin of distinction between classes, this algorithm performs reasonably well. It is more effective in high-dimensional areas and uses less memory.
Decision Tree (DT)
Decision Tree (DT), like SVM, is a Supervised Learning Approach. It is a tree-like structured classifier in which internal nodes contain dataset attributes, all the branches indicate decision rules, and every leaf node represents the outcome. It is named a decision tree because, like a tree, it begins with the parent or root node and then develops on subsequent branches to form a tree-like structure. Decision trees need less effort for data processing than other methods. This technique does not demand data normalization or scaling. This approach is very obvious and simple to communicate to both technical teams and stakeholders.
K-Nearest Neighbor(KNN)
The K-Nearest Neighbour method is the most basic Machine Learning technique. This model suggests resemblance between the new instances and preexisting instances and places the new instances in the category which is most compatible with the existing categories. It is a non-parametric method, that implies it makes no assumptions regarding information. It is also known as a lazy learner algorithm because it does not instantly learn from the training set; alternatively, it accumulates the information and then takes the appropriate action on it during categorization.
Naive Bayes(NB)
The Naive Bayes method was created on the basis of Bayesian statistics and is used to solve classification issues. This Classifier is one of the simplest and most successful Classification algorithms, assisting in the development of rapid machine learning models capable of making rapid forecasts. It is a probabilistic classifier, which implies that it predicts based on the likelihood of an object.
Random Forest(RF)
Random forest classifier is made up of a large number of individual decision trees that work together. The basic goal of this strategy is to train numerous "primitive" decision trees before applying the generic algorithm to every tree. This approach generates a random sample with substitutions from the training dataset and fits trees to these choices on a continuous basis. Random Forest then integrates the projections from every tree and calculates the final outcome depending on the predictions that won the most votes.

In this study, the combination of these two models, AFI and AGI, was used to predict a person's gender from a source image.
Figure 2. Entire process of the suggested approach according to the embodiment of the present invention.
Table 1 shows the reliability analysis of the two hybrid models for gender identification.
Table 1: Performance Evaluation of Predictive Models for Gender Classification
Methods Accuracy Specificity Sensitivity
AFI+AGI [Support Vector Machine] 93.44% 0.9615 0.8984
AFI+AGI [K-Nearest Neighbour] 90.63% 0.9375 0.8750
AFI+AGI [Naïve Bayes] 87.76% 0.9200 0.8333
AFI+AGI [Decision Tree] 94.49 % 0.9720 0.9185
AFI+AGI [Random Forest] 97.12 % 0.9851 0.9568

In the case of the AGI model, different classifiers such as SVM, NB, K-NN, DT, RF are employed. It is a well-known fact that SVM, NB, K-NN, DT are single learner models where the DT model achieves the most promising accuracy of 94.49%. As a consequence, another model RF is employed where multiple DT models are assembled. Finally, it can be concluded that application of multiple DT in RF has boosted the efficiency (97.12%).
Figure 3. Overall experimental result of automated gender prediction according to the embodiment of the present invention.
[012] Without further elaboration, the foregoing will so fully illustrate my invention, that others may, by applying current of future knowledge, readily adapt the same for use under various conditions of service. It should also be realized by those skilled in the art that such equivalent constructions do not depart from the spirit and scope of the invention.

Advantages over the prior art
[013] An intelligent gender identification method based on facial images using compound modules proposed by the present invention has the following advantages over the prior art:
1. High Accuracy: The system achieves a high classification accuracy of 97.12% using the random forest algorithm, ensuring reliable gender detection from both images and video files.
2. Real-Time Processing: The technology can process facial data in real-time, making it suitable for dynamic environments like security systems, surveillance, and live interactions.
3. Hybrid Approach: By combining machine learning, deep learning, and temporal and spectral domain analysis, the system benefits from the strengths of multiple advanced techniques, enhancing its overall performance and accuracy.
4. Versatility: It can classify gender not only from static images but also from video files, expanding its applicability across different media types and use cases, including crime scene investigations and visual tracking systems.
5. Environmentally and User-Friendly: The system is designed to be adaptable to various environments and user-friendly, making it accessible for a wide range of applications, from security to customer service.
6. Soft Biometrics Integration: Gender classification, as part of the soft biometrics category, adds a layer of information to broader identity recognition systems, improving overall system efficiency and functionality in biometric applications.
7. Scalability: The approach can be scaled to accommodate large databases and integrated into existing facial recognition systems, allowing it to serve various industries with minimal disruption.
8. Enhanced Security and Monitoring: The technology's ability to accurately identify gender in real-time contributes to enhanced security monitoring, personal safety, and identity verification, making it useful in diverse security and surveillance applications.
[014] In the preceding specification, the invention has been described with reference to specific exemplary embodiments thereof. It will be evident that various modifications and changes may be made thereunto without departing from the broader spirit and scope of the invention as set forth in the claims that follow. The specification and drawings are accordingly to be regarded in an illustrative rather than restrictive sense. Therefore, the aim in the appended claims is to cover all such changes and modifications as fall within the true spirit and scope of the invention. The matter set forth in the foregoing description and accompanying drawings is offered by way of illustration only and not as a limitation. The actual scope of the invention is intended to be defined in the following claims when viewed in their proper perspective based on the prior art.

, Claims:We claim:
1. An intelligent gender identification method based on facial images using compound modules comprising of:
- Detecting a face from an input image using a face detection algorithm;
- Identifying key facial landmarks from the detected face;
- Extracting features from the detected face based on the identified facial landmarks using frequency domain analysis;
- Constructing a feature vector from the extracted features; and
- Classifying the gender of the person based on the constructed feature vector.
2. The method of claim 1, wherein the face detection algorithm is the Viola-Jones algorithm.
3. The method of claim 1, wherein the facial landmarks include the eyes, nose, mouth, and eyebrows.
4. A system for automated gender classification, comprising:
- A face detection module configured to detect a human face from an input image;
- A facial landmark identification module configured to identify key facial parts from the detected face;
- A feature extraction module configured to extract facial features from the identified landmarks using Fast Fourier Transform (FFT) and Discrete Cosine Transform (DCT); and
- A gender classification module configured to classify the gender of the person based on the extracted features.
5. The system of claim 4, wherein the gender classification module employs a Random Forest classifier, which integrates predictions from multiple decision trees to determine gender.
6. The system of claim 4, wherein the feature extraction module employs a combination of Fast Fourier Transform (FFT) and Discrete Cosine Transform (DCT) to convert facial image data from the spatial domain to the frequency domain.
7. A computer-implemented method for gender identification, comprising:
- Receiving an input image of a person;
- Detecting a face within the input image using a face detection algorithm;
- Identifying facial landmarks, including the eyes, nose, mouth, and eyebrows, using a landmark detection algorithm;
- Converting the extracted features from spatial domain to frequency domain using Fast Fourier Transform (FFT) and Discrete Cosine Transform (DCT);
- Generating a feature vector based on the frequency domain features; and
- Classifying the gender of the person using a machine learning algorithm selected from the group consisting of Support Vector Machine (SVM), Naive Bayes (NB), K-Nearest Neighbors (KNN), Decision Tree (DT), and Random Forest (RF).
8. The method of claim 7, wherein the machine learning algorithm used for gender classification is Random Forest, and the Random Forest algorithm includes multiple Decision Trees that are trained on the generated feature vector.
9. A non-transitory computer-readable medium storing instructions for performing the method of claim 1, wherein the instructions, when executed by a processor, cause the processor to detect a face, extract facial features, construct a feature vector, and classify the gender of the person.
10. The method of claim 1, wherein the input image is selected from the group consisting of a still image and a video frame, and wherein the method further comprises identifying and classifying gender from video files in real-time.

Documents