Enhance the AI Virtual System Accuracy with Novel Hand Gesture Recognition Algorithm Comparing to Convolutional Neural Network

: The objective of this study is to enhance the precision of AI virtual systems by implementing Novel Hand Gesture Recognition techniques in comparison to Convolutional Neural Network. Materials and Methods: To recognise hand motions, a Convolutional Neural Network with distinct training and testing stages is utilized. The average Gpower for the test is between 0.05 and 0.85, or around 85%. Sample size is determined as 27,455 for each group using G Power 3.1 software (G Power setting parameters: α=0.05 and power=0.85). Results and Discussion: Novel Hand gesture recognition 92.60% identifies between objects and boosts the observed accuracy with a statistically non-significant value of p=0.123 (p>0.05) in comparison to the convolutional neural network's 88.59%. Conclusion: Comparison of the Novel Hand gesture Recognition algorithm and Convolutional Neural Network in terms of performance that shows Hand gesture recognition has 91.62% with better accuracy.


INTRODUCTION
In order to robustly model visual characteristics, this research proposes a novel hand gesture recognition algorithm that makes use of machine learning techniques.Each gesture is represented by the algorithm as a collection of fixed sub-gestures, each of which consists of a succession of frames with locally coherent context.The sub-gesture modeling technique that has been presented may be useful in the area of Human-Computer Interaction (HCI) [1], particularly in systems that recognise gestures.The suggested method makes use of filler and gesture completion models to precisely identify the edges of gestures.The Novel Hand Gesture Recognition algorithm is a new algorithm created to increase the reliability and accuracy of hand gesture recognition systems [2].The proposed algorithm can recognise hand gestures efficiently and effectively using machine learning, improving human-computer interaction in various industries [3].The advancement of human gesture recognition (HGR) has had a significant positive impact on the research of human-computer interaction.Our suggested solution makes use of real-time hand segmentation, feature extraction, and gesture recognition methods.We use the precise realtime findings of the Camshift and HSV color models for hand tracking and segmentation.The application of evolutionary algorithms to gesture recognition improves recognition accuracy [4].Because HGR algorithms streamline procedures and boost efficiency, they have transformed not only HCI but the entire industry.The HGR algorithm-based AI virtual system has shown enormous potential in areas including gaming, healthcare, and the auto industry, where it has enhanced control and machine interaction.[5] This study demonstrates the use of six static hand gestures and eight moving hand gestures to control a computer.The system utilizes AI virtual systems that incorporate Novel Hand Gesture Recognition (HGR) algorithms and Machine Learning (ML) techniques to accurately recognize and respond to the gestures [6].The use of such systems is becoming increasingly prevalent in a wide range of industries, such as gaming, healthcare, and automotive, where they can improve human-computer interaction and enhance user experience There are three primary steps: recognising a hand shape, tracking the detected hand, and translating the information into the required instruction [7].The project's overarching objective is to bridge the gap between humans and machines via the use of AI-powered gesture recognition that can share information better [8].Hand gesture recognition allows for more intuitive and immersive interactions, enabling users to perform actions such as grabbing, pointing, and manipulating virtual objects with their hands, which mimics real-world interactions [9].This technology is also being used in sign language recognition, where hand gestures are used as a form of communication for individuals with hearing impairment.Hand gesture recognition in AI virtual systems also finds applications in gaming, healthcare, and human-computer interaction, where users can control applications, navigate menus, or perform actions using hand gestures, providing a more natural and seamless user experience [10].Additionally, hand gesture recognition can be used in security systems for authentication purposes, where users can perform specific hand gestures as a form of biometric identification.
Overall, hand gesture recognition in AI virtual systems has the potential to revolutionize the way humans interact with virtual environments and applications, enhancing user experience and enabling more intuitive and natural interactions.The current system's identification of research gaps was found to be incorrect.In order to improve classification accuracy, this research proposes to implement a novel hand gesture recognition algorithm and evaluate its performance against that of a convolutional neural network (CNN) .The GPower program was used to compare two controllers and calculate the appropriate sample size.Two groups were selected for comparison, and ten samples were taken from each group for the experiment.Technical analysis tools were used to implement two algorithms: the CNN algorithm and the Novel Hand Gesture Recognition algorithm.The sample size for each group was calculated to be 27,455 using the GPower 3.1 program with a power of 0.85 and a significance level of α=0.05.

MATERIALS AND METHODS
Python and OpenCV were used to implement the task, and Windows 10 with an Intel Core i5 CPU and 4GB of RAM was used for testing.The 64-bit system was used.For code implementation, Python was used as the programming language (Fronteddu et al. 2022).Picture, objects, distance, frequency, modulation, amplitude, volume, and dependent components were among the factors taken into account for the study.The qualities or elements of the hand gesture that are utilized to categorize it are often referred to as the independent variables.As the code executes, the dataset is being processed in the background to generate the output and ensure accuracy.
This study employed an open source dataset for Hand Gesture Recognition using images, which consists of more than 2000 images and has a total size of around 256MB.The dataset was obtained from an open source repository and was divided into training and validation sets.

Hand Gesture Recognition Algorithm
The term "hand gesture recognition algorithms" refers to a group of approaches that analyze and interpret hand gestures acquired in images or videos by using computer vision and machine learning techniques.These algorithms aim to recognize specific hand gestures by collecting and processing relevant information from the input data, enabling them to classify the hand movements into predefined categories.Through the use of advanced algorithms, these systems can accurately identify and track hand movements, which can be useful in a variety of applications, as gaming, human-computer interaction, and sign language recognition.

Statistical Analysis
The statistical analysis for Hand Gesture Recognition was conducted using IBM SPSS version 21.The variables considered for analysis included picture objects, separation, frequency, amplitude, volume, modulation, and dependent components.The qualities or elements of the hand gesture that are utilized to categorize it are often referred to as the independent variables.The categorization or label given to a certain gesture based on its attributes is often the dependent variable in a hand gesture recognition algorithm.The dependent variables for both procedures were things and pictures.Separate T-test analyses were performed to determine the accuracy of both methods.The accuracy of the Hand Gesture Recognition algorithm and Convolutional Neural Network (CNN) was evaluated, and the results were compared using mean accuracy [11].The independent t-test analyses were carried out to compute the accuracy of both methods.The results showed that the proposed Hand Gesture Recognition algorithm approach gave an accuracy of 92.62%, , 04022 (2024) E3S Web of Conferences https://doi.org/10.1051/e3sconf/202449104022491 ICECS'24 which was insignificantly better in precise classification when compared to the CNN algorithm, which had an accuracy of 88.59%.

RESULTS
This study produced two algorithms, Novel Hand Gesture Recognition and Convolutional Neural Network (CNN), which were assessed using a sample size of 27,455 based on their accuracy and loss measurements.Table 2 shows the anticipated accuracy and loss of the CNN algorithm, whereas Table 1 lists the precision and reduction in hand gesture identification forecasts.The CNN algorithm's mean accuracy was determined to be 88.59%, but the mean accuracy for the Novel Hand Gesture Recognition was found to be 92.62%.The findings imply that when it comes to hand gesture recognition, Novel Hand Gesture Recognition is more precise than CNN.Table 3 summarizes the mean accuracy results for the two approaches, with hand gesture recognition having a lower mean value of 92.62% and a standard deviation of 2. With a p=0.123 (p>0.05), the T-test for independence between hand gesture recognition and CNN revealed that there is a significant difference between the two methods in Table 4.The accuracy of HGR and CNN is displayed in Table 5.A graphic depiction of the number of photographs according to their labels is shown in Figure 1.The history of Accuracy and Loss is depicted in Figure 2. The comparison between the mean accuracy and loss of hand gesture recognition and CNN is shown graphically in Figure 3.
The Innovative Hand Gesture Recognition method has a mean accuracy of 92.6030, a standard deviation of 4.69803, and a mean standard error of 1.48565.The Convolutional Neural Network technique, on the other hand, has a mean accuracy of 88.5950, a standard deviation of 6.27997, and a standard error mean of 1.98590.In terms of loss, the convolutional neural network's mean value is 11.4050, with a standard deviation of 6.27997 and a standard error mean of 1.98590, compared to the mean value for hand gesture recognition of 7.3970, 4.69803, and 1.48565.The T-test for independence between Novel Hand Gesture Recognition and Support Vector Machines yielded a p=0.123 (p>0.05) based on a sample size of 27,455, demonstrating no discernible difference between the two methods.However, compared to Support Vector Machines, which had a mean accuracy of 88.5950 and a standard deviation of 6.27997, the Novel Hand Gesture Recognition technique performed more accurately, with a mean accuracy of 92.6030 and a standard deviation of 4.69803.The New Hand Gesture Recognition approach is hence more reliable and accurate at identifying hand gestures.Moreover, the hand gesture recognition standard deviation is reduced, indicating more reliable outcomes.Based on the findings of this study, it is suggested that hand gesture recognition is a more effective approach than convolutional neural networks for accurately recognizing static hand gestures.The statistical analysis revealed a statistically non-significant difference between the two techniques with a p=0.123 (p>0.05).The accuracy of hand gesture recognition was reported to be 92.60%,while the accuracy of convolutional neural networks was found to be 88.59%.The high accuracy rate of hand gesture recognition indicates its potential for practical applications in fields such as human-computer interaction and sign language recognition.This study highlights the effectiveness of machine learning techniques for the task of static hand gesture recognition.
In this paper, Machine learning techniques, particularly convolutional neural network (CNN), are commonly used in hand gesture recognition (HGR) studies(Ikram and Liu 2021).Both HGR and CNN involve extracting features and classifying data into predefined categories [12].However, different studies use varying feature extraction and classification methods, leading to varying accuracy levels (Nayan, Ghosh, and Pradhan 2022).Similarities are Both are used for classification tasks and can learn features automatically from data.Dissimilarities are HGR is a specific task that deals with recognizing hand gestures, while CNN is a more general algorithm that can be applied to a wide range of tasks.Also, HGR often uses simpler architectures than CNNs [13].In the construction industry, hand gestures are used by crane operators to communicate with ground crew to position heavy equipment safely.In the automotive industry, hand gestures are used by assembly line workers to coordinate tasks and maintain workflow efficiency.In the mining industry, hand gestures are used by drill operators to communicate with team members about drilling depth and direction.In the aerospace industry, hand gestures are used by air traffic controllers to direct ground crew during aircraft movements on the tarmac.For instance, studies have integrated surface electromyography (sEMG) signals with CNN to achieve around 89% accuracy rates.
Novel segmentation methods have also achieved 79.6% accuracy rates [14].Despite differences, both HGR and CNN have potential to improve accuracy through integration of other technologies.Factors such as lighting and gesture complexity can affect system accuracy, and continued research can lead to more reliable and accurate HGR systems [15].One of the study's drawbacks is how long it takes to train a hand gesture detection system, especially when working with large datasets.Industry observers consider the use of ML to construct AI virtual systems using HGR algorithms to be a game-changer.According to the study's future goals, the system should be extended to include more things while needing less time to train the data set.

CONCLUSION
Two distinct methods for hand gesture recognition are compared in this research study: Novel Hand Gesture Recognition and Convolutional Neural Networks (CNNs).The accuracy scores of these two methods are evaluated using a particular dataset.The accuracy score of the Novel Hand Gesture Recognition approach is found to be 92.60%.This approach likely involves using innovative techniques or algorithms that are different from traditional methods of hand gesture recognition.These novel techniques may have been developed or adapted specifically for the dataset used in this research, leading to a high , 04022 (2024) E3S Web of Conferences https://doi.org/10.1051/e3sconf/202449104022491 ICECS'24 accuracy score.On the other hand, the accuracy score of the Convolutional Neural Network approach is 88.59%.CNNs are a type of deep learning model that are commonly used for image recognition tasks, including hand gesture recognition.The accuracy score of 88.59% suggests that the CNN has performed well in recognizing hand gestures from the dataset, but it is slightly lower than the accuracy of the Novel Hand Gesture Recognition approach.

Fig 2 :
Fig 2: The History of Accuracy and loss of CNN and HGR according to Training and Validation.

Fig 3 .
Fig 3.Bar Chart representing the comparison of Mean Accuracy of HGR and CNN algorithms.MeanAccuracy of HGR is better than CNN and standard deviation is slightly better for HGR than CNN .X-Axis: HGR vs CNN, Y-Axis: Mean Accuracy of detection +/-2 SD One unique type of network design for deep learning is convolutional neural networks.learning methods used in image recognition and other jobs that need the processing of pixel input.Machine learning-based filler and gesture completion models are used in Human-Computer Interaction to recognize gesture boundaries and improve gesture spotting.Convolutional neural networks are often used for a range of computer vision tasks, including image classification, face recognition, object detection, etc (Convolutional Neural Networks).

Table 1 .
Accuracy and Loss Analysis of Hand Gesture Recognition for a sample size of 27,455.

Table 2 .
Accuracy and Loss Analysis of Convolutional Neural Network(CNN) for a sample size of 27,455.

Table 4 .
Group Statistical Analysis of Novel Hand gesture recognition and Convolutional Neural Network.

Table 5 .
Independent Sample T-test: Novel Hand gesture recognition is insignificantly better than Convolutional Neural Network with p = 0.123 (p>0.05).

Table 6 .
Comparison of the Novel Hand gesture recognition and Convolutional Neural Network with their accuracy