Abstract
Face detection, recognition and gender estimation are one of the most significant research areas in computer vision, not only because of the
challenging nature of faces as an object but also due to the countless
applications that require the application of face detection, tracking and
recognition. Although many significant types of research on face detection,
recognition and gender estimation problems have done in the last few years
separately, there is no particular research on face detection, recognition and
gender estimation together from a real-time video for person identification. So,
we feel that these types of significant research are still needed to work. The
main contributions of our paper are divided into three parts, namely face
detection, recognition and gender estimation for person identification. In our
research work, we use Local Binary pattern Histogram (LBPH) method and
Convolution Neural Network (CNN) to extract the facial features of face
images whose computational complexity is very low. By calculating the Local
Binary Patterns Histogram (LBPH) neighborhood pixels and Convolution
levels, we extract effective facial feature to realize face recognition and gender
estimation. We show the experimental results using these methods to recognize
face and gender for person identification. CNN increase the calculating speed
of testing real-time video and also improve the recognition rate. By using
LPBH, we get 63% accuracy on average where CNN gives 99.88% training
accuracy for face recognition-1 and 96.88% accurate for gender estimation-1
and 100% training accuracy for face recognition-2 and 93.38% training
accuracy for gender estimation-2. However, Convolution Neural Networks
(CNN) learns fast and predict efficiently.
challenging nature of faces as an object but also due to the countless
applications that require the application of face detection, tracking and
recognition. Although many significant types of research on face detection,
recognition and gender estimation problems have done in the last few years
separately, there is no particular research on face detection, recognition and
gender estimation together from a real-time video for person identification. So,
we feel that these types of significant research are still needed to work. The
main contributions of our paper are divided into three parts, namely face
detection, recognition and gender estimation for person identification. In our
research work, we use Local Binary pattern Histogram (LBPH) method and
Convolution Neural Network (CNN) to extract the facial features of face
images whose computational complexity is very low. By calculating the Local
Binary Patterns Histogram (LBPH) neighborhood pixels and Convolution
levels, we extract effective facial feature to realize face recognition and gender
estimation. We show the experimental results using these methods to recognize
face and gender for person identification. CNN increase the calculating speed
of testing real-time video and also improve the recognition rate. By using
LPBH, we get 63% accuracy on average where CNN gives 99.88% training
accuracy for face recognition-1 and 96.88% accurate for gender estimation-1
and 100% training accuracy for face recognition-2 and 93.38% training
accuracy for gender estimation-2. However, Convolution Neural Networks
(CNN) learns fast and predict efficiently.
Original language | English |
---|---|
Pages (from-to) | 395-415 |
Number of pages | 21 |
Journal | Journal of Computer Science |
Volume | 15 |
Issue number | 3 |
DOIs | |
Publication status | Published - 2019 |