el 21(22): e4

Research Article

Lip language identification via Wavelet entropy and K-nearest neighbor algorithm

Download59 downloads
  • @ARTICLE{10.4108/eai.11-8-2021.170669,
        author={Ran Wang and Yifan Cui and Xinyu Gao and Wei Chen and Mingbo Hu and Qian Li and Jiahui Wei and XianWei Jiang},
        title={Lip language identification via Wavelet entropy and K-nearest neighbor algorithm},
        journal={EAI Endorsed Transactions on e-Learning},
        volume={7},
        number={22},
        publisher={EAI},
        journal_a={EL},
        year={2021},
        month={8},
        keywords={Lip language identification, Wavelet entropy, 𝐾𝐾-nearest neighbor, Wavelet transform, K-fold cross validation},
        doi={10.4108/eai.11-8-2021.170669}
    }
    
  • Ran Wang
    Yifan Cui
    Xinyu Gao
    Wei Chen
    Mingbo Hu
    Qian Li
    Jiahui Wei
    XianWei Jiang
    Year: 2021
    Lip language identification via Wavelet entropy and K-nearest neighbor algorithm
    EL
    EAI
    DOI: 10.4108/eai.11-8-2021.170669
Ran Wang1, Yifan Cui1, Xinyu Gao1, Wei Chen1, Mingbo Hu1, Qian Li1, Jiahui Wei1, XianWei Jiang1,*
  • 1: School of Mathematics and Information Science, Nanjing Normal University of Special Education, Nanjing 210038, China
*Contact email: jxw@njts.edu.cn

Abstract

INTRODUCTION: Image processing technology is widely used in lip recognition, which can automatically detect and analyse the unstable shape of human lips.

OBJECTIVES: In this paper, we propose a new algorithm using Wavelet entropy (WE) and K-nearest neighbor (KNN) improves the accuracy of lip recognition.

METHODS: At present, the two most commonly used technologies are wavelet transform and 𝐾𝐾-nearest neighbor algorithm. Wavelet transform is a set of image descriptors, and the 𝐾𝐾-nearest neighbor algorithm has high accuracy. After a large
number of experiments, we propose a lip recognition method based on Wavelet entropy and 𝐾𝐾-nearest neighbor, which combines Wavelet entropy, 𝐾𝐾-nearest neighbor and K-fold cross validation.

RESULTS: This method reduces the calculation time and improves the training speed. The best result of the experiment improves the accuracy to 80.08%.

CONCLUSION: Therefore, our algorithm is superior to other state-of-the-art approaches of lip recognition.