WOA-TLBO: Whale optimization algorithm with Teaching-learning-based optimization for global optimization and facial emotion recognition
Introduction
Facial expression is one of the top significant features of human emotion recognition. Even though much research on facial emotion recognition has been conducted over the years, it still faces numerous challenges. The challenges are primarily due to interpersonal differences, the subtlety of facial expressions, posture, and illumination, etc. Facial Emotion Recognition [1], [2] has become an increasingly important research area that involves many applications such as human–computer interaction (HCI), driver safety, healthcare, deceive detection and security, etc. Currently, FER in pictures has intrigued rising consideration, which is for models further convoluted due to low-resolution faces and backgrounds. To categorize facial emotions like fear, disgust, happiness, sadness, anger, neutrality, and surprise has the foremost objective. Facial expression (FE) is the most convincing, frequent, and consistent way for humans to communicate their intentions and expressive conditions.
Based on Paul and Friesen’s research work, most FER approaches are proposed [3]. The available tasks are commonly characterized into two groups: geometric-based methods and appearance-based methods [4]. The geometric-based process typically needs to spot the facial feature point whose similar activities can bolster apprehend the expressional features. The appearance-based way uses the texture modality and explores expressional variances in pixel space. Robust emotion taxonomy relies closely on dominant facial representation. However, it is still a thought-provoking undertaking for categorizing substantial discriminative facial features that could constitute all emotional characteristics due to the nuance and changeability of facial expressions [5], [6]. In [7], authors anticipated LBP (Local Binary Pattern) centered FER methods, in that the features remained discriminate and henceforth carried other critical statistics. But, the technique is not appropriate for partly occluded pictures. In [8], the authors delivered a FER procedure centered on the Support-vector machine (SVM) that be erect effective for the trade-off attention in accuracy and computation complexity. In [9], the authors presented a MultiSV (Multi-class Support Vector) classifier appropriate for the many illumination properties. The disadvantage of the tactic is that the noise will sternly influence accurateness. In [10], the authors planned a countenance acknowledgment scheme with a unique bedded cascade biological procedure for optimal feature choice. In [11], the authors suggested a countenance perception scheme with hvnLBP grounded feature extraction, mGA-embedded PSO-centered feature development, and numerous classifier-founded emotion recognition. By motivated by these works, in this paper we have proposed the effective method of FER using WOA-TLBO based MultiSVNN.
This paper uses real-time facial expression recognition to deliver beneficially and enhanced discriminative facial representation to deal with such problems. In assessment with other feature selection approaches, EC (Evolutionary Computational) systems demonstrate dominant global search competencies and have been extensively acknowledged as proficient methods for feature excerpts [12]. The humpback whales’ social activities inspire one of the different EC procedures, the whale optimization algorithm (WOA) [13]. It has remained widely used for optimizing the features with the benefits of a rapid convergence rate and low computational cost. But, standard WOA inclines to meet before the usual time and, as a result, be trapped in local optimum [13]. As an outcome, a WOA is embedded with the Teaching Learning Based Optimization Algorithm (TLBO) [14], named as WOA-TLBO is proposed in this paper.
The main contributions of this paper consist of: (1) ELDP (Efficient Local Directional Pattern) and SLDP (Scatter LDP) feature: At the first stage, the ST (Scattering Transform) and the LBP are applied to the facial image. Then the LDP is used for the facial image in the second stage. The yield attained from the Local Directional Pattern and ST is exposed to the EX-OR process that harvests an SLDP picture in the third step. The SLDP and ELDP trait engendered from the picture using the LDP and ST such that they contemporaneous the vigorous attributes for recognition. The investigational outcomes designate that it encompasses added squeamish power than the regular LDP for FER. (2) WOA-TLBO-centered Multi-SVNN: The proposed WOA-TLBO procedure for refinement of the optimum hefts of Multi-SVNN to advance recognition efficiency and accurateness.
The anticipated FER arrangement comprises three phases: (a) feature extraction, (b) feature optimization, and (c) emotion recognition. The information that brings the facial phantasmagorias is utilized for the acknowledgment method, and the facial picture is dynamic to pull out the features. The SIFT (Scale-Invariant Feature Transform) [15], ELDP, and SLDP descriptors are used to extract the features. The recognition procedure uses the features created using the descriptors SIFT, SLDP, and ELDP. Then the classifier that classifies the picture dependent on the facial highlights of the image. The scheme is assessed with three facial emotion datasets, i.e., the Amsterdam Dynamic Facial Expression Sets (ADFES) [16], Cohn–Kanade AU-Coded Expression Datasets [17], and The Japanese Female Facial Expression (JAFFE) Datasets [18]. The experimental outcomes designate that the proposed system outdoes state-of-the-art optimization methods.
The structure of the rest of the paper is organized in the following way. We discuss the related work in Section 2. In Sections 3 Whale optimization algorithm, 4 Teaching-learning-based optimization (TLBO), we introduced the basics of WOA and TLBO. The proposed WOA-TLBO algorithm is offered in Section 5. The evaluation of WOA-TLBO using benchmark test functions is covered in Section 6. The suggested facial emotion recognition technique is described in Section 7. Finally, the conclusion is provided in Section 8.
Section snippets
Related work
The literature suggests that emotion recognition from humanoid face images can be used in emotional computing, and the facial emotion recognition survey can be found in [19], [20]. In [19], the authors examined advanced resolutions such as face registration, dimensionality devaluation, illustration, and identification by disassembling pipelines into principal components. In [21], the Active Appearance Model (AAM) appropriate technique used for FER. For validating and cataloging deformable
Whale optimization algorithm
Seyedali Mirjalili and Andrew Lewis proposed a population-based metaheuristic algorithm, namely, WOA [38], which mimics humpback whales’ social behavior. After attaining prey’s position, humpback whales will dive twelve meters down and trail spiral with bubbles upward to the surface. Fig. 1 shows the Bubble-net hunting behavior of humpback whales.
Encircling prey, Bubble-net attacking phase (exploitation stage), and the hunt for prey are the three steps that the role takes up to date
Teaching-learning-based optimization (TLBO)
Teaching-learning-based optimization (TLBO) [14] is a meta-heuristic technique inspired by the classroom situation’s learning and teaching procedure. The population is considered a group of students, and the best student is selected as the teacher by the fitness evaluation. The process of the TLBO method is separated into two stages, Teacher Stage and Learner Stage.
Proposed algorithm
In this segment, the particulars of the proposed WOA-TLBO procedure is shown. The WOA is concentrated on global exploration, and the TLBO is focused on fast convergence speed. Thus the complication is in what way to create the pre-eminent of the benefits of TLBO and WOA. To resolve this delinquent, centered on the suitability values of entities, the population is separated into two halves. The nastiest half of the population and the most elegant half of the population are applied via WOA and
Experiment preparation and Parameters settings for the compared algorithms
To authenticate the WOA-TLBO algorithm’s enactment, an enormous amount of experimentations are prepared on benchmark test functions. The benchmark test functions used for experiments are exhibited in Table 1. Through many times sovereign turns, we can achieve the Standard Deviation (SD), the Average (Ave), worst value (worst), and Median (Med) values of each method on every benchmark test function. Usually, the SD value represents the steadiness, and the Ave value signifies the optimization
The proposed facial emotion recognition system
Facial Emotion recognition (FER) is hot research in current intelligent systems [8], [51]. The method is persistently used in human–computer interaction (HCI), remote medical service, and autonomous vehicles. Many researchers are committed to the study of FER and have predicted several functional approaches [52], [53].
The paper aims at an improved approach to recognizing human appearances based on their routine developments, for which the expression of the individual is used as part of the mien
Conclusion
Facial emotion recognition (FER) finds several real-life applications such as education, security, human–robot interaction, medicine, etc. The standard methodologies neglect to separate the articulation absolutely as the couplings rely upon the segments’ endeavor like the mouth, ear, etc. Thus, to affirm the powerful affirmation, the anticipated SIFT+SLDP+ELDP+WOA-TLBO-MultiSVNN is locked in. The features of the facial picture are taken out utilizing the SIFT, SLDP, and ELDP descriptors. The
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
References (60)
- et al.
IWOA: An improved whale optimization algorithm for optimization problems
J. Comput. Des. Eng.
(2019) - et al.
Teaching–learning-based optimization: A novel method for constrained mechanical design optimization problems
Comput. Des.
(2011) - et al.
Deep spatial–temporal feature fusion for facial expression recognition in static images
Pattern Recognit. Lett.
(2019) - et al.
Facial expression recognition with convolutional neural networks: Coping with few data and the training sample order
Pattern Recognit.
(2017) - et al.
The whale optimization algorithm
Adv. Eng. Softw.
(2016) - et al.
Snap-drift cuckoo search: A novel cuckoo search optimization algorithm
Appl. Soft Comput. J.
(2017) - et al.
Salp Swarm Algorithm: A bio-inspired optimizer for engineering design problems
Adv. Eng. Softw.
(2017) SCA: A Sine Cosine Algorithm for solving optimization problems
Knowl.-Based Syst.
(2016)Moth-flame optimization algorithm: A novel nature-inspired heuristic paradigm
Knowl.-Based Syst.
(2015)- et al.
Grasshopper optimisation algorithm: Theory and application
Adv. Eng. Softw.
(2017)
Hybridizing sine cosine algorithm with differential evolution for global optimization and object tracking
Appl. Soft Comput. J.
A synergy of the sine-cosine algorithm and particle swarm optimizer for improved global optimization and object tracking
Swarm Evol. Comput.
A practical tutorial on the use of nonparametric statistical tests as a methodology for comparing evolutionary and swarm intelligence algorithms
Swarm Evol. Comput.
Intelligent facial emotion recognition based on stationary wavelet entropy and jaya algorithm
Neurocomputing
Multi-cue fusion for emotion recognition in the wild
Neurocomputing
Eigenvalue decay: A new method for neural network regularization
Neurocomputing
Feature selection algorithm based on bare bones particle swarm optimization
Neurocomputing
Enhanced leader PSO (ELPSO): A new PSO variant for solving global optimisation problems
Appl. Soft Comput. J.
Face recognition using transform domain feature extraction and PSO-based feature selection
Appl. Soft Comput. J.
HEPSO: High exploration particle swarm optimization
Inf. Sci. (Ny).
ImageNet classification with deep convolutional neural networks
Adv. Neural Inf. Process. Syst.
A survey on human face expression recognition techniques
J. King Saud Univ. - Comput. Inf. Sci.
Constants across cultures in the face and emotion
Local directional ternary pattern for facial expression recognition
IEEE Trans. Image Process.
Boosting local gabor binary patterns for gender recognition
Proc. - Int. Conf. Nat. Comput.
Automatic facial expression recognition: Feature extraction and selection
Signal, Imag. Video Process.
Extraction of informative regions of a face for facial expression recognition
IET Comput. Vis.
Facial expression recognition using a combination of multiple facial features and support vector machine
Soft Comput.
Facial appearance and texture feature-based robust facial expression recognition framework for sentiment knowledge discovery
Clust. Comput.
Intelligent facial emotion recognition using a layered encoding cascade optimization model
Appl. Soft Comput. J.
Cited by (38)
Study of flow stress in Mg-Gd-Y-Nd-Zr alloys based on IWOA-BPNN model
2024, Journal of Materials Research and TechnologyComprehensive learning TLBO with recursive precedence-based solution construction and multilevel local search for the linear ordering problem
2024, Expert Systems with ApplicationsAn efficient Optimization State-based Coyote Optimization Algorithm and its applications
2023, Applied Soft ComputingRanking teaching–learning-based optimization algorithm to estimate the parameters of solar models
2023, Engineering Applications of Artificial Intelligence