Whale Optimization Algorithm-Driven Gait Recognition Model with ROI Extraction and Hybrid Classification Approach

Problem Definition

The existing literature on human identification using gait features reveals a number of limitations that hinder the accuracy and efficacy of current models. One major issue is the lack of implementation of segmentation techniques for extracting the Region of Interest (ROI) from images, resulting in reduced accuracy. Furthermore, the absence of feature extraction and selection techniques in standard models leads to the dimensionality curse, further degrading the performance of the models. Another noteworthy point is the limited use of hybrid models, which have the potential to significantly improve the efficiency and efficacy of human identification models. It is evident from these findings that there is a critical need for a new and improved human identification model that addresses these limitations and enhances the overall performance of gait-based recognition systems.

Objective

The objective of this study is to address the limitations identified in the existing literature on human identification using gait features. These limitations include the lack of segmentation techniques for extracting the Region of Interest (ROI) from images, the absence of feature extraction and selection techniques leading to the dimensionality curse, and the limited use of hybrid models. The proposed work aims to overcome these shortcomings by implementing PCA and GLCM techniques for feature extraction and classification, using a tree-based model tuned with the WOA optimization algorithm. By extracting the ROI, selecting important features, and utilizing hybrid models, the objective is to improve the accuracy and efficiency of human identification models.

Proposed Work

From the above literatures, it is observed that over the years, a significant number of approaches have been proposed by various researchers for identifying humans using gait features. However, these models undergo through a number of limitations that degrade their accuracy rate. Majority of the researchers didn’t use any segmentation technique in their models for extracting the Region of Interest (ROI) from images, which reduces accuracy of the models. In addition to this, no feature extraction and selection technique was implemented in standard models which leads to dimensionality curse and degrades the efficacy of the model. Moreover, it was also observed that not majority of work has been done using hybrid models that can really increase the efficiency and efficacy of human identification models.

Keeping these findings in mind, a new and improved human identification model must be proposed for overcoming these shortcomings. In this project, we have implemented PCA and GLCM technique for feature extraction and classification of human gait using a tree-based model tuned with WOA optimization algorithm. The objective is to address the existing limitations identified in the literature and improve the accuracy and efficiency of human identification models. In order to achieve the desired results, the proposed gait based human identification model collects the necessary information from OULP-CIVI-A database. Since the images present contain a lot of unnecessary information that increases the complexity of the system if passed directly to classifiers, it is important to extract the Region of Interest (ROI) from images so that only the important and informative part of the image is passed down to classifiers.

By doing so, all the unnecessary data present in the image is removed and only the informative part of the image is obtained, which in turn reduces the complexity and processing time of the proposed model. This is followed up by the feature extraction process wherein important and crucial features like skewness, Kurtosis, Root mean Square (RMS) and GLCM features like Energy, contrast, correlation, and Homogeneity features are extracted. This aids in reducing the dimensionality of the dataset which further decreases the complexity of the proposed Human identification method. Finally, the classification process is initiated wherein the featured images are passed to SVM, ANN, and WOA-DT classifiers to analyze their performance in terms of various parameters. Each step in the proposed work has been carefully chosen to address the identified limitations and improve the overall efficiency and accuracy of the human identification model.

Application Area for Industry

This project can be applied across various industrial sectors to enhance security measures through improved human identification models. The proposed solutions address challenges faced by industries such as surveillance, access control, and biometric authentication by implementing segmentation techniques to extract the Region of Interest (ROI) from images. By using feature extraction and selection techniques, the dimensionality curse is reduced, leading to more accurate and efficient human identification models. The utilization of hybrid models further increases the efficacy of the system by combining different classifiers like SVM, ANN, and WOA-DT. Implementing these solutions can result in improved security measures, reduced processing time, and enhanced accuracy rates within different industrial domains.

Application Area for Academics

The proposed project on gait-based human identification can greatly enrich academic research, education, and training in the fields of computer vision, biometrics, and machine learning. By addressing the limitations identified in existing models, the project offers a novel approach to accurately identify individuals based on their gait features. Educationally, this project can serve as a valuable resource for students pursuing research in the area of biometric identification systems. It provides a practical example of how segmentation techniques, feature extraction, and selection methods can be applied to enhance the accuracy and efficiency of human identification models. This hands-on experience with state-of-the-art algorithms and classifiers like SVM, ANN, and WOA-DT can offer valuable insights into the field of machine learning and data analysis.

Researchers in the specific domain of biometrics and computer vision can utilize the code and literature of this project to build upon existing knowledge and further advance the field. Moreover, MTech students and PhD scholars can leverage the proposed methods and algorithms to explore innovative research methods, conduct simulations, and analyze data within educational settings. The potential applications of this project are vast, ranging from improving security systems to enhancing surveillance technology. By incorporating hybrid models and advanced algorithms, the proposed human identification model has the potential to revolutionize the way individuals are identified based on their unique gait patterns. In conclusion, the proposed project has the potential to significantly contribute to academic research, education, and training by offering a comprehensive approach to human identification through gait analysis.

The future scope of this project includes further refining the model, exploring additional classifiers, and expanding the dataset to improve the accuracy and reliability of the system.

Algorithms Used

The proposed gait-based human identification model utilizes various algorithms to improve accuracy and efficiency. The first step involves extracting the Region of Interest (ROI) from images to eliminate unnecessary information, thereby reducing system complexity. Next, features such as skewness, kurtosis, RMS, and GLCM features are extracted to reduce dataset dimensionality. Finally, the featured images are classified using SVM, ANN, and WOA-DT classifiers to evaluate performance in terms of various parameters. By integrating segmentation, feature extraction, and classification techniques, the model aims to address limitations present in existing human identification models and enhance overall efficacy and efficiency.

Keywords

SEO-optimized keywords: human identification, gait features, segmentation technique, Region of Interest, feature extraction, selection technique, dimensionality curse, hybrid models, efficiency, efficacy, OULP-CIVI-A database, unnecessary information, complexity, processing time, skewness, Kurtosis, Root mean Square, GLCM features, Energy, contrast, correlation, Homogeneity, dataset dimensionality, SVM classifier, ANN classifier, WOA-DT classifier, PCA, tree-based model, WOA optimization algorithm.

SEO Tags

problem definition, human identification models, gait features, segmentation technique, region of interest, feature extraction, dimensionality curse, efficacy, hybrid models, human identification model, OULP-CIVI-A database, image processing, feature extraction process, skewness, kurtosis, root mean square, GLCM features, energy, contrast, correlation, homogeneity, classification process, SVM classifier, ANN classifier, WOA-DT classifier, PCA, GLCM, tree-based model, WOA optimization algorithm