Classification of Sign Language Numbers Using the CNN Method
Abstract
Abstrak Berkomunikasi merupakan kebutuhan semua individu karena setiap individu harus berkomunikasi dengan lingkungan. Berkomnikasi juga membuat seseorang mendapat informasi sehingga dapat dijadikan acuan untuk beradaptasi. penggunaan bahasa verbal dengan berbicara mengeluar suara adalah cara komunikasi individu, namun hal itu tidak dapat dilakukan saat berkomunikasi dengan individu yang memilki keterbatasan dalam mendengar. Keterbatasan tersebut membuat diperlukan cara komunikasi lain yaitu melalui bahasa isyarat. Bahasa isyarat banyak jenisnya salah satunya bahasa isyarat menggunakan tangan membentuk huruf atau angka. Bahasa isyarat terdapat standar, standar yang cukup terkenal adalah standar American Sign Language (ASL). Masih banyak yang sulit mengenal bahasa isyarat, maka solusinya adalah membuat sistem untuk klasifikasi bahasa isyarat. Penelitian ini akan membuat sistem machine learning untuk pengenalan angka bahasa isyarat standar American Sign Language (ASL) serta menerapkan preprocessing untuk optimalisasi hasil. Hasil penelitian ini adalah melakukan perbandingan metode preproscessing yang diterapkan pada sistem Convlutional neural network arsitektur mobilenetv2. Hasil akhir penelitian kombinasi metode preprocessing Grayscale, HSV, Global Threshold menghasilkan akuarasi pengenalan terbaik yaitu 97%.
Abstract Communicating is a need for all individuals because an individual must communicate with the environment. Communicating also enables someone to obtain information so that it can serve as a reference for adaptation. The use of spoken language while speaking out of a voice is an individual means of communication, but it cannot be applied when communicating with persons with hearing limitations. These limitations require another way of communication, namely through sign language. There are many kinds of ASL, one of which is ASL using hands to form letters or numbers. Standard popular Sign language is the American Sign Language (ASL) standard. Many still people difficult to recognize sign language, so a solution is to create a system for sign language classification. This research will create a machine learning system for number recognition in American standard sign language. Sign Language (ASL) as well as applying preprocessing to optimize results. The result of this research is to compare the recognition accuracy of the scenarios of different preprocessing methods applied in the Convolutional neural network system architecture MobileNetV2. The final result of this research is the combination of Grayscale, HSV, and Global Threshold preprocessing method yielding the best recognition accuracy of 97%.
References
[2] X. Luo, X. Qin, Z. Wu, F. Yang, M. Wang, and J. Shang, “Sediment Classification of Small-Size Seabed Acoustic Images Using Convolutional Neural Networks,” IEEE Access, vol. 7, pp. 98331–98339, 2019, doi: 10.1109/ACCESS.2019.2927366.
[3] V. Borate, S. Patange, V. Vede, and O. Kale, “An Image Classification Based on CNN Approach For Plant Leaf Disease Detection,” vol. 4, no. 6, p. 3, 2018.
[4] S. Z. M. Zaki, M. Asyraf Zulkifley, M. Mohd Stofa, N. A. M. Kamari, and N. Ayuni Mohamed, “Classification of tomato leaf diseases using MobileNet v2,” IJ-AI, vol. 9, no. 2, p. 290, Jun. 2020, doi: 10.11591/ijai.v9.i2.pp290-296.
[5] I. K. G. Darma Putra, R. Fauzi, D. Witarsyah, and I. P. D. Jayantha Putra, “Classification of Tomato Plants Diseases Using Convolutional Neural Network,” International Journal on Advanced Science, Engineering and Information Technology, vol. 10, no. 5, p. 1821, Oct. 2020, doi: 10.18517/ijaseit.10.5.11665.
[6] Computer Science and Engineering Department, National Institute of Technology Manipur, Imphal, 795001, India, R. Meitram, and P. Choudhary, “Palm Vein Recognition Based on 2D Gabor Filter and Artificial Neural Network,” JAIT, vol. 9, no. 3, pp. 68–72, 2018, doi: 10.12720/jait.9.3.68-72.
[7] M. Hurroo and M. E. Walizad, “Sign Language Recognition System using Convolutional Neural Network and Computer,” International Journal of Engineering Research, vol. 9, no. 12, p. 6.
[8] A. Mavi, “A New Dataset and Proposed Convolutional Neural Network Architecture for Classification of American Sign Language Digits,” p. 5.
[9] J. W. Gotama Putra, Pengenalan Konsep Pembelajaran Mesin dan Deep Learning, 1.4. 2020. [Online]. Available: https://www.researchgate.net/publication/323700644_Pengenalan_Pembelajaran_Mesin_dan_Deep_Learning
[10] Md. M. Kabir, A. Q. Ohi, Md. S. Rahman, and M. F. Mridha, “An Evolution of CNN Object Classifiers on Low-Resolution Images,” in 2020 IEEE 17th International Conference on Smart Communities: Improving Quality of Life Using ICT, IoT and AI (HONET), Charlotte, NC, USA, Dec. 2020, pp. 209–213. doi: 10.1109/HONET50430.2020.9322661.
[11] F. Sultana, A. Sufian, and P. Dutta, “Advancements in Image Classification using Convolutional Neural Network,” 2018 Fourth International Conference on Research in Computational Intelligence and Communication Networks (ICRCICN), pp. 122–129, Nov. 2018, doi: 10.1109/ICRCICN.2018.8718718.
[12] M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, and L.-C. Chen, “MobileNetV2: Inverted Residuals and Linear Bottlenecks,” arXiv:1801.04381 [cs], Mar. 2019, Accessed: Feb. 02, 2021. [Online]. Available: http://arxiv.org/abs/1801.04381
[13] R. C. N. Santi, S. Pd, and M. Kom, “Mengubah Citra Berwarna Menjadi GrayScale dan Citra biner,” vol. 16, p. 6, 2011.
[14] K. Bhargavi and S. Jyothi, “A Survey on Threshold Based Segmentation Technique in Image Processing,” vol. 3, no. 12, p. 7, 2014.
[15] N. P. Sutramiani, Ik. G. Darmaputra, and M. Sudarma, “Local Adaptive Thresholding Pada Preprocessing Citra Lontar Aksara Bali,” JTE, vol. 14, no. 1, Jun. 2015, doi: 10.24843/MITE.2015.v14i01p06.
