Academia.eduAcademia.edu

Figure 7 – uploaded by Journal of Computer Science IJCSIS

See full PDFdownloadDownload figure
Visual Results under the complex background and light variation:

Figure 7 Visual Results under the complex background and light variation:

Related Figures (13)

Order histogram features. Linear neural network are used fot recognition in parallel mode. A. Noise Removal In subjected paper we used MDWMF technique for removal of RVIN. In this technique they used second order derivative for noise detection. JCSIS) International Journal of Computer Science and Information Security, Vol. 18, No. 04, April 202(
Order histogram features. Linear neural network are used fot recognition in parallel mode. A. Noise Removal In subjected paper we used MDWMF technique for removal of RVIN. In this technique they used second order derivative for noise detection. JCSIS) International Journal of Computer Science and Information Security, Vol. 18, No. 04, April 202(
D. Morphological operation Morphology is the integrity of different image processing operation. These techniques deal with the shape of image. Structure elements are used in all morphological operations. In proposed technique we use only erosion and dilation. We used 5*5 structure element for erosion.
D. Morphological operation Morphology is the integrity of different image processing operation. These techniques deal with the shape of image. Structure elements are used in all morphological operations. In proposed technique we use only erosion and dilation. We used 5*5 structure element for erosion.
FEATURE EXTRACTION (1ST ORDER HISTOGRAM) In equation (5) Structure element B is applied on the image Then image is converted to binary and canny edge detector is used to get finer edges of the segmented region. We used Ist Order histogram feature to get statistical values of the segmented region. In proposed technique we extract the following features:
FEATURE EXTRACTION (1ST ORDER HISTOGRAM) In equation (5) Structure element B is applied on the image Then image is converted to binary and canny edge detector is used to get finer edges of the segmented region. We used Ist Order histogram feature to get statistical values of the segmented region. In proposed technique we extract the following features:
Figure 3: Dilation Structure Element
Figure 3: Dilation Structure Element
Fig 4: One against All Technique In one against all technique, the NN algorithm is trained for every class separately. Consider the subjective class 1 and then the remaining 4 assign label 0. In proposed technique NN is trained for each class separately and for testing result are taken from each trained neural network. Result are stored in array and taken the index of maximum number which identify the related class.
Fig 4: One against All Technique In one against all technique, the NN algorithm is trained for every class separately. Consider the subjective class 1 and then the remaining 4 assign label 0. In proposed technique NN is trained for each class separately and for testing result are taken from each trained neural network. Result are stored in array and taken the index of maximum number which identify the related class.
20 images with size 256*256 are taken for the examination of technique. We applied 10 fold techniques for analyzing the accuracy of our proposed technique. We use 10-fold validation techniques for learning and testing. In 100 images we use 90 images for leaning and 10images for testing to examine the accuracy. In second iteration another 10images are selected for classification and the previous selected images are combine with training data. We have 100 image datasets against each class. MDWMEF [8] are used due to highest PSNR in existing noise removal techniques, and image enhancement technique which helped to achieve better segmentation.
20 images with size 256*256 are taken for the examination of technique. We applied 10 fold techniques for analyzing the accuracy of our proposed technique. We use 10-fold validation techniques for learning and testing. In 100 images we use 90 images for leaning and 10images for testing to examine the accuracy. In second iteration another 10images are selected for classification and the previous selected images are combine with training data. We have 100 image datasets against each class. MDWMEF [8] are used due to highest PSNR in existing noise removal techniques, and image enhancement technique which helped to achieve better segmentation.
Table 4: Comparison of the results Because we use one against all technique so we divided the gesture group in 5 combinations.
Table 4: Comparison of the results Because we use one against all technique so we divided the gesture group in 5 combinations.
Recognition Accuracy:
Recognition Accuracy:
Table 2: Recognition Result of proposed Method
Table 2: Recognition Result of proposed Method
Aashni et. al., hand gesture Simple thresho the background correctly under 2017 proposed a technique to recognize the for Human Computer Interaction. They use Iding method to segment the hand region from . Drawback of this technique is it not working poor lighting and with complex back ground. S 0 our proposed method covers these drawbacks and achieve improved result under poor luminance and in complex background. T he visual results of segmentation accuracy are shown above table 2. They use Contour Extraction and convex hull as a featur e. Haar Cascade Classifier is used to recognize the gesture. For validation of proposed technique, the same dataset is used and compare the accuracy results in the following method. Consider the accuracy of segmented image.
Aashni et. al., hand gesture Simple thresho the background correctly under 2017 proposed a technique to recognize the for Human Computer Interaction. They use Iding method to segment the hand region from . Drawback of this technique is it not working poor lighting and with complex back ground. S 0 our proposed method covers these drawbacks and achieve improved result under poor luminance and in complex background. T he visual results of segmentation accuracy are shown above table 2. They use Contour Extraction and convex hull as a featur e. Haar Cascade Classifier is used to recognize the gesture. For validation of proposed technique, the same dataset is used and compare the accuracy results in the following method. Consider the accuracy of segmented image.
Table 3: Comparison of the results
Table 3: Comparison of the results
Figure 7: Result Comparison Graph Conclusion
Figure 7: Result Comparison Graph Conclusion
Related topics:
Information SystemsComputer ScienceInformation ScienceInformation TechnologyInformation SecurityComputer SecurityHand Gesture Recognition SystemVision-based hand gesture recognition
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved