• Proceedings of the Korea Information Processing Society Conference
• /
• 2014.04a
• /
• pp.781-784
• /
• 2014

Face recognition under controlled settings, such as limited viewpoint and illumination change, can achieve good performance nowadays. However, real world application for face recognition is still challenging. In this paper, we use Affine SIFT to detect affine invariant local descriptors for face recognition under large viewpoint change. Affine SIFT is an extension of SIFT algorithm. SIFT algorithm is scale and rotation invariant, which is powerful for small viewpoint changes in face recognition, but it fails when large viewpoint change exists. In our scheme, Affine SIFT is used for both gallery face and probe face, which generates a series of different viewpoints using affine transformation. Therefore, Affine SIFT allows viewpoint difference between gallery face and probe face. Experiment results show our framework achieves better recognition accuracy than SIFT algorithm on FERET database.

• Journal of Korea Multimedia Society
• /
• v.25 no.4
• /
• pp.568-574
• /
• 2022

In this paper, direction normalization and CNN deep learning are used to develop a more reliable license plate recognition system. The existing license plate recognition system consists of three main modules: license plate detection module, character segmentation module, and character recognition module. The proposed system minimizes recognition error by adding a direction normalization module when a detected license plate is inclined. Experimental results show the superiority of the proposed method in comparison to the previous system.

• The Journal of the Acoustical Society of Korea
• /
• v.22 no.5
• /
• pp.388-402
• /
• 2003

In this paper, we carried out the word, 4 continuous digits. continuous, and task-independent word recognition experiments to verify the effectiveness of the re-defined phoneme-likely units (PLUs) for the phonetic decision tree based HM-Net (Hidden Markov Network) context-dependent (CD) acoustic modeling in Korean appropriately. In case of the 48 PLUs, the phonemes /ㅂ/, /ㄷ/, /ㄱ/ are separated by initial sound, medial vowel, final consonant, and the consonants /ㄹ/, /ㅈ/, /ㅎ/ are also separated by initial sound, final consonant according to the position of syllable, word, and sentence, respectively. In this paper. therefore, we re-define the 39 PLUs by unifying the one phoneme in the separated initial sound, medial vowel, and final consonant of the 48 PLUs to construct the CD acoustic models effectively. Through the experimental results using the re-defined 39 PLUs, in word recognition experiments with the context-independent (CI) acoustic models, the 48 PLUs has an average of 7.06%, higher recognition accuracy than the 39 PLUs used. But in the speaker-independent word recognition experiments with the CD acoustic models, the 39 PLUs has an average of 0.61% better recognition accuracy than the 48 PLUs used. In the 4 continuous digits recognition experiments with the liaison phenomena. the 39 PLUs has also an average of 6.55% higher recognition accuracy. And then, in continuous speech recognition experiments, the 39 PLUs has an average of 15.08% better recognition accuracy than the 48 PLUs used too. Finally, though the 48, 39 PLUs have the lower recognition accuracy, the 39 PLUs has an average of 1.17% higher recognition characteristic than the 48 PLUs used in the task-independent word recognition experiments according to the unknown contextual factor. Through the above experiments, we verified the effectiveness of the re-defined 39 PLUs compared to the 48PLUs to construct the CD acoustic models in this paper.

• Journal of Korea Multimedia Society
• /
• v.11 no.6
• /
• pp.787-795
• /
• 2008

Human recognition on camera is an interesting topic in computer vision. While fingerprint and face recognition have been become common, gait is considered as a new biometric feature for distance recognition. In this paper, we propose a gait recognition algorithm based on the knee angle, 2 feet distance, walking velocity and head direction of a person who appear in camera view on one gait cycle. The background subtraction method firstly use for binary moving object extraction and then base on it we continue detect the leg region, head region and get gait features (leg angle, leg swing amplitude). Another feature, walking speed, also can be detected after a gait cycle finished. And then, we compute the errors between calculated features and stored features for recognition. This method gives good results when we performed testing using indoor and outdoor landscape in both lateral, oblique view.

• The Journal of the Acoustical Society of Korea
• /
• v.21 no.2
• /
• pp.160-166
• /
• 2002

In this paper, we implement a new vocabulary-independent speech recognition system that uses CV, VCCV, VC recognition unit. Since these recognition units are extracted in the trowel region of syllable, the segmentation is easy and robust. And in the case of not existing VCCV unit, the units are replaced by combining VC and CV semi-syllable model. Clustering of vowel group and applying combination rule to the substitution model in the case of not existing of VCCV model lead to 5.2% recognition performance improvement from 90.4% (Model A) to 95.6% (Model C) in the first candidate. The recognition results that is 98.8% recognition rate in the second candidate confirm the effectiveness of the proposed method.

• Journal of Korea Society of Digital Industry and Information Management
• /
• v.15 no.4
• /
• pp.149-157
• /
• 2019

In case of reading passport using a smart phone in contrast with a dedicated passport reading system, MRZ(Machine Readable Zone) character recognition can be hard when the character strokes were broken, touched or blurred according to the lighting condition, and the position and size of MRZ character lines were varied due to the camera distance and angle. In this paper, the effective recognition algorithm of the passport MRZ information using a combined neural network recognizer of CNN(Convolutional Neural Network) and ANN( Artificial Neural Network), is proposed under the various sized and skewed passport images. The MRZ line detection using connected component analysis algorithm and the skew correction using perspective transform algorithm are also designed in order to achieve effective character segmentation results. Each of the MRZ field recognition results is verified by using five check digits for deciding whether retrying the recognition process of passport MRZ information or not. After we implement the proposed recognition algorithm of passport MRZ information, the excellent recognition performance of the passport MRZ information was obtained in the experimental results for PC off-line mode and smart phone on-line mode.

• Journal of Korea Multimedia Society
• /
• v.11 no.12
• /
• pp.1676-1687
• /
• 2008

In this paper, we proposed a new multi-view human recognition method based on face and gait features detection algorithm. For getting the position of moving object, we used the different of two consecutive frames. And then, base on the extracted object, the first important characteristic, walking direction, will be determined by using the contour of head and shoulder region. If this individual appears in camera with frontal direction, we will use the face features for recognition. The face detection technique is based on the combination of skin color and Haar-like feature whereas eigen-images and PCA are used in the recognition stage. In the other case, if the walking direction is frontal view, gait features will be used. To evaluate the effect of this proposed and compare with another method, we also present some simulation results which are performed in indoor and outdoor environment. Experimental result shows that the proposed algorithm has better recognition efficiency than the conventional sing]e view recognition method.

• The Journal of the Acoustical Society of Korea
• /
• v.16 no.4
• /
• pp.42-48
• /
• 1997

This paper has apparaised ability of speaker recognition and speech recognition using correlation dimension and Lyapunov dimension. In this method, speech was regarded the cahos that the random signal is appeared in determinisitic raising system. we deduced exact correlation dimension and Lyapunov dimension with searching important orbit from AR model power spectrum when reconstruct strange attractor using Taken's embedding theory. We considered a usefulness of speech recognition and speaker recognition using correlation dimension and Lyapunov dimension that characterized reconstruction attractor. As a result of consideration, which were of use more the speaker recognition than speech recognition, and in case of speaker recognition using Lyapunov dimension were much recognition rate more than speaker recognitions using correlation dimension.

• Journal of Korea Multimedia Society
• /
• v.21 no.5
• /
• pp.626-637
• /
• 2018

Human exchanges information not only through words, but also through body gesture or hand gesture. And they can be used to build effective interfaces in mobile, virtual reality, and augmented reality. The past 2D gesture recognition research had information loss caused by projecting 3D information in 2D. Since the recognition of the gesture in 3D is higher than 2D space in terms of recognition range, the complexity of gesture recognition increases. In this paper, we proposed a real-time gesture recognition deep learning model and application in 3D space using deep learning technique. First, in order to recognize the gesture in the 3D space, the data collection is performed using the unity game engine to construct and acquire data. Second, input vector normalization for learning 3D gesture recognition model is processed based on deep learning. Thirdly, the SELU(Scaled Exponential Linear Unit) function is applied to the neural network's active function for faster learning and better recognition performance. The proposed system is expected to be applicable to various fields such as rehabilitation cares, game applications, and virtual reality.

• The Journal of the Acoustical Society of Korea
• /
• v.10 no.3
• /
• pp.31-36
• /
• 1991

We propose an efficient pre-classification algorithm extracting candidate words to reduce the recognition time in a large-vocabulary recognition system and also propose the use of spectral and temporal smoothing of the observation probability to improve its classification performance. The proposed algorithm computes the coarse likelihood score for each word in a lexicon using the observation probabilities of speech spectra and duration information of recognition units. With the proposed approach we could reduce the computational amount by 74% with slight degradation of recognition accuracy in 1160-word recognition system based on the phoneme-level HMM. Also, we observed that the proposed coarse likelihood score computation algorithm is a good estimator of the likelihood score computed by the Viterbi algorithm.