WO2006057475A1 - Face detection and authentication apparatus and method - Google Patents
Face detection and authentication apparatus and method Download PDFInfo
- Publication number
- WO2006057475A1 WO2006057475A1 PCT/KR2004/003480 KR2004003480W WO2006057475A1 WO 2006057475 A1 WO2006057475 A1 WO 2006057475A1 KR 2004003480 W KR2004003480 W KR 2004003480W WO 2006057475 A1 WO2006057475 A1 WO 2006057475A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- lace
- data
- image
- detection
- authentication
- Prior art date
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 107
- 238000000034 method Methods 0.000 title claims abstract description 61
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 86
- 238000012545 processing Methods 0.000 claims description 7
- 238000010586 diagram Methods 0.000 description 17
- 239000000284 extract Substances 0.000 description 8
- 238000000605 extraction Methods 0.000 description 8
- 238000012549 training Methods 0.000 description 8
- 238000011156 evaluation Methods 0.000 description 5
- 210000001061 forehead Anatomy 0.000 description 4
- 238000013459 approach Methods 0.000 description 3
- 230000001419 dependent effect Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000006854 communication Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 210000005069 ears Anatomy 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000001815 facial effect Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000012821 model calculation Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
- G06V40/171—Local features and components; Facial parts ; Occluding parts, e.g. glasses; Geometrical relationships
Definitions
- the AdaBoost learning algorithm is used for enhancing lace recognition rates.
- the AdaBoost learning algorithm which is simple but efficient compared to other boost algorithms, is used for more precisely differentiating a lacial area from a non ⁇ acial area and boosting the probability of an area of an image determined to be a lacial area.
- locations of prototypes of Haar-like features can render more detailed lacial characteristics especially at a higher stage of classification.
- Haar-like features obtained using the AdaBoost learning algorithm are classified in stages, as shown in FIG. 2. Referring to FIG.
- an area of an image of one frame is determined as a lacial area using a group of trained lace images 300 while gradually decreasing the size of the image in a pyramid manner.
- a plurality of candidate areas for the lacial area of the image are generated in the process of recovering the size of the image, and an average of the candidate areas is output.
- Facial areas of trained lace images are detected and then differentiated from non ⁇ acial areas of the trained lace images using lace classifiers. Thereafter, lace authentication is performed by comparing the detected fecial area with the trained lace images using an HMM.
- a feature extraction unit 520 extracts features O , O , ..., O from the
- Table 1 shows trained location information and critical values of prototypes of
- FIG. 7 is a diagram illustrating lace authentication results obtained using the lace detection and authentication method according to an exemplary embodiment of the present invention.
- 48 lace images of 8 people (6 lace images per person) and a trained database of 24) lace images of 4) people (6 lace images per person) were used in experiments for testing the lace detection and authentication apparatus according to the exemplary embodiment of the present invention.
- the size of the forehead in lace images of people who ⁇ iled to be authenticated is larger than in lace images of people who were successfiilly verified, as shown in FlG. 7. Because of a larger forehead, the eyes in the lace images of those who f ⁇ iled to be authenticated deviate from expected locations when segmenting the corresponding lace images in a 2D HMM-based training process.
- FIG. 9 is a diagram illustrating another example of the lace detection and au- thentication method according to the exemplary embodiment of the present invention that uses a WPS 80.
- the WPS 83 automatically extracts &ce data from an image to be authenticated using a lace detection algorithm stored therein and wirelessly transmits the extracted lace data using a JPEG codec to a server 90.
- the server 90 stores a database of lace images and a lace authentication algorithm.
- the server 90 includes a lace authentication unit, which authenticates the image to be authenticated by comparing the extracted lace data with lace data stored therein.
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- General Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Collating Specific Patterns (AREA)
Abstract
A face detection and authentication apparatus and method, which decide whether to authenticate face data detected from an input face image using face detection algorithms by comparing the detected face data with face data previously stored, are provided. The face detection and authentication apparatus and method can detect face data from a moving picture and can authenticate the detected face data using a wearable personal station (WPS) and a server or using only the WPS and thus can provide efficient solutions for identifying and authenticating people, providing them with a log-on function, and holding video conferences.
Description
Description FACE DETECTION AND AUTHENTICATION
APPARATUS AND METHOD
Technical Field
[1] The present invention relates to a lace detection and authentication apparatus and method, and more particularly, to a lace detection and authentication apparatus and method that detect a lace image from a moving picture using a personal portable terminal and a server that communicate with each other or using only the personal portable terminal, compare the lace image with a lace image registered in a database, and determine whether to authenticate the lace image based on the comparison results.
Background Art
[2] In the processing and interpretation of images, human laces are a critical lactor of visual distinction and identification of people. Since the early 1990s, various analysis tools for lace recognition and fecial expression interpretation have been developed. Recently, MPEG-7 based lace descriptors for lace searching and identification have been suggested. MPEG-7 based lace descriptors can help to more quickly and more ef¬ ficiently search a database for a lace image that is most likely to be a match for an input lace image than conventional lace recognition algorithms.
[3] Face recognition is a technique of identifying a person by comparing the person's still image or moving picture with a given database of lace images. Face recognition acquires biometric information from a person to be verified in a less intrusive manner than other biometric recognition techniques, such as fingerprint recognition, without requiring the person to directly contact a recognition system with part of his or her body. However, the lace is subjected to various changes due to illumination and poses and is more sensitive to the surroundings than other parts of the body. Thus, recognition rates of lace recognition systems are lower than recognition rates of other biometric recognition systems.
[4] Examples of conventional lace detection and recognition techniques are disclosed in Korean Patent Laid-open No. 10-1005-29985 entitled 'Method of Identifying Person by Processing Person's Face Image', Korean Patent Laid-open No. 10-1998-20738 entitled 'Face Feature Extraction System and Method', Korean Patent Laid-open No. 10-2000-6D745 entitled 'Real-time Face Tracking Method Using Face Color Models and Ellipsoid Approximation Model', and Korean Patent Laid-open No. 10-2001-87487 entitled 'Face Recognizing Method and Apparatus Using Dimensional
Space Transformation of Gabor Filtering Response'. However, these patented lace recognition techniques have a clear limitation in achieving high lace recognition rates and are considered as having few practical applications. Disclosure of Invention
Technical Problem
[5] The present invention provides a lace detection and authentication apparatus and method, which capture a moving picture input to a personal portable terminal, wirelessly transmit the captured moving picture to a server, detect lace data from the moving picture using one of a plurality of lace detection algorithms that can be most quickly applied to the moving picture, authenticate the detected lace data by comparing in real time the detected lace data with lace data stored in the server, and output the authentication results to the personal portable terminal.
[6] The present invention also provides a lace detection and authentication apparatus and method, which detect lace data from a moving picture stored in a personal portable terminal using one of a plurality of lace detection algorithms that can be most quickly applied to the moving picture, wirelessly transmit the detected lace data to a server, au¬ thenticate the detected lace data by comparing in real time the detected lace data with lace data stored in a database of the server using a lace authentication algorithm stored in the server, and output the authentication results to the personal portable terminal.
[7] The present invention also provides a lace detection and authentication apparatus and method, which detect lace data from a moving picture stored in a personal portable terminal using one of a plurality of lace detection algorithms that can be most quickly applied to the moving picture, wirelessly transmit parameters of features of the detected lace data to a server, and authenticate the detected lace data based on the parameters of the features of the detected lace data by comparing in real time the detected lace data with lace data stored in a database of the server using a lace au¬ thentication algorithm stored in the server.
[8] The present invention also provides a lace detection and authentication apparatus and method, which detect lace data from a moving picture stored in a personal portable terminal using one of a plurality of lace detection algorithms that can be most quickly applied to the moving picture, extract parameters of features of the detected lace data, and authenticate the detected lace data in the personal portable terminal based on the parameters of the features of the detected lace data by comparing in real time the detected lace data with lace data stored in a database of the personal portable terminal.
Technical Solution
[9] According to an aspect of the present invention, there is provided a lace detection and authentication apparatus. The lace detection and authentication apparatus includes: a personal portable terminal, which processes a lace image of a moving picture input thereto, stores the processed lace image, and wirelessly transmits the processed lace image; and a server, which comprises a storage unit storing a lace database, lace detection algorithms, and lace authentication algorithms and a lace authentication unit authenticating lace data extracted from the lace image that is wirelessly transmitted by the personal portable terminal by comparing the extracted lace data with lace data stored in the lace database and then outputting the authentication results.
[10] According to another aspect of the present invention, there is provided a lace detection and authentication method. The lace detection and authentication method includes: storing a lace database, lace detection algorithms, and lace authentication algorithms in a server; processing a lace image of a moving picture input to a personal portable terminal, storing the processed lace image, and wirelessly transmitting the processed lace image from the personal portable terminal to the server; and au¬ thenticating lace data extracted from the lace image that is wirelessly transmitted by the personal portable terminal by comparing the extracted lace data with lace data stored in the lace database stored in the server and then outputting the authentication results.
[11] According to another aspect of the present invention, there is provided a lace detection and authentication apparatus. The lace detection and authentication apparatus includes: a personal portable terminal, which comprises a storage unit storing lace detection algorithms and a lace detection unit detecting lace data from a moving picture input thereto; and a server, which comprises a storage unit storing a lace database and lace authentication algorithms and a lace authentication unit au¬ thenticating the detected lace data by comparing the detected lace data with lace data stored in the lace database and then outputting the authentication results.
[12] According to another aspect of the present invention, there is provided a lace detection and authentication method. The lace detection and authentication method includes: storing a lace database and lace authentication algorithms in a server; detecting lace data from a moving picture input to a personal portable terminal using lace detection algorithms stored in the personal portable terminal and wirelessly transmitting the detected lace data from the personal portable terminal to the server; and authenticating the detected lace data that is wirelessly transmitted by the personal
portable terminal by comparing the detected lace data with lace data stored in the lace database stored in the server and then outputting the authentication results.
[13] According to another aspect of the present invention, there is provided a lace detection and authentication apparatus. The lace detection and authentication apparatus includes: a personal portable terminal, which comprises a storage unit storing lace data and lace detection algorithms and a lace detection unit detecting lace data from a moving picture input thereto and wirelessly transmitting parameters of features of the detected lace data; and a server, which comprises a storage unit storing a lace database and lace authentication algorithms and a lace authentication unit authenticating the detected lace data based on the parameters of the features that is wirelessly transmitted by the personal portable terminal by comparing the detected lace data with lace data stored in the lace database and then outputting the authentication results.
[14] According to another aspect of the present invention, there is provided a lace detection and authentication method. The lace detection and authentication method involves: storing a lace database and lace authentication algorithms in a server; detecting lace data from a moving picture input to a personal portable terminal using lace detection algorithms stored in the personal portable terminal and wirelessly transmitting parameters of features of the detected lace data from the personal portable terminal to the server; and authenticating the detected lace data based on the parameters of the features that is wirelessly transmitted by the personal portable terminal by comparing the detected lace data with lace data stored in the lace database stored in the server and then outputting the authentication results.
[15] According to another aspect of the present invention, there is provided a lace detection and authentication apparatus. The lace detection and authentication apparatus includes: a storage unit, which stores lace data in a personal portable terminal; a lace detection unit, which detects lace data from a moving picture input to the personal portable terminal and outputs parameters of features of the detected lace data; and a lace authentication unit, which authenticates the detected lace data based on the parameters of the features by comparing the detected lace data with lace data stored in a lace database and then outputs the authentication results.
[16] According to another aspect of the present invention, there is provided a lace detection and authentication method. The lace detection and authentication method involves: storing lace data in a personal portable terminal; detecting lace data from a moving picture input to the personal portable terminal and outputting parameters of features of the detected lace data; and authenticating the detected lace data based on
the parameters of the features by comparing the detected lace data with lace data stored in a lace database and then outputting the authentication results.
[17] Face data may be detected from a moving picture by laying a window over a føcial area of the moving picture, obtaining Haar-like feature values of the lacial area, and then classifying and storing the Haar-like feature values using prototypes of Haar-like features and the AdaBoost learning algorithm.
[18] A lace image may be authenticated by generating a hidden Markov model (HMM) based on features of lace data detected from the lace image using an HMM algorithm and then sequentially applying the HMM model to the lace image in the order from the top to the bottom of the lace image.
[19] In the HMM algorithm, a two dimensional (2D) HMM may be generated through the expansion of each block of the lace image by dividing each superstate into a plurality of states, calculating the probabilities of the states, and calculating the probability of a superstate based on the probabilities of the states into which the superstate is divided.
Description of Drawings
[20] FIG. 1 is a diagram illustrating prototypes of Haar-like features applied to the present invention;
[21] FIG. 2 is a diagram illustrating stages of classification using the AdaBoost learning algorithm applied to the present invention;
[22] FIG. 3 is a flowchart of a lace detection method using Haar-like features, according to an exemplary embodiment of the present invention;
[23] FIG. 4 is a diagram illustrating a process of generating lace data using a hidden
Markov model (HMM) algorithm and a process of calculating probabilities of su¬ perstates in lace authentication according to an exemplary embodiment of the present invention;
[24] FIG. 5 is a block diagram illustrating a process of training lace data for lace detection and authentication and a process of generating a database of lace data in the lace detection and authentication method according to an exemplary embodiment of the present invention;
[25] FIG. 6 is a block diagram of a lace detection and authentication apparatus according to an exemplary embodiment of the present invention;
[26] FIG. 7 is a diagram illustrating lace authentication results obtained using the lace detection and authentication method according to an exemplary embodiment of the present invention;
[27] FlG. 8 is a diagram illustrating an example of the lace detection and authentication method according to the exemplary embodiment of the present invention;
[28] FlG. 9 is a diagram illustrating another example of the lace detection and au¬ thentication method according to the exemplary embodiment of the present invention;
[29] FlG. 10 is a diagram illustrating another example of the lace detection and au¬ thentication method according to the exemplary embodiment of the present invention; and
[30] FlG. 11 is a diagram illustrating another example of the lace detection and au¬ thentication method according to the exemplary embodiment of the present invention.
Best Mode
[31] The present invention will now be described more fully with reference to the ac¬ companying drawings in which exemplary embodiments of the invention are shown.
[32] In the lace recognition field, a Haar-like feature-based lace detection algorithm achieves short training time and enhanced processing speed and thus is expected to be very usefiil in lace detection based on moving pictures. A lace recognition system can be realized based on the Haar-like feature-based lace detection algorithm and a hidden Markov model (HMM) algorithm. The lace recognition system extracts lace data from a lace image to be verified, compares the extracted lace data with trained lace images stored in a database, and verifies the lace image as one of the trained lace images that is most likely to be a match for the extracted lace data.
[33] Face detection and identification rates of the lace recognition system are determined through trainings, and thus the performance of the lace recognition system is dependent on the number of trainings and the efficiency of each of the trainings. The lace recognition system only requires a person to be verified to show his or her lace to a camera, which is very simple and convenient compared to other recognition systems.
[34] In the present invention, a lace detection and authentication apparatus and method are realized based on Haar-like features and the HMM algorithm. Specifically, in the present invention, lace authentication is performed using an HMM in order to enhance lace recognition rates. In an HMM approach, a lace image is divided into segments and then processed in units of the segments. Face data that is needed to construct input image data is extracted from a lace image using Haar-like features (reference: P. Viola and M. J. Jones, 'Robust Real-time Object Detection,' Technical Report Series, Compaq Cambridge Research Laboratory, CRL 2001/01, Feb. 2001). In other words, lace authentication is performed by comparing the extracted lace data with a database of trained lace images using the HMM algorithm and then outputting one of the trained
lace images that is most likely to be a match for the extracted lace data as the comparison results. Thus, it is possible to economically enhance lace recognition rates.
[35] Face detection techniques applied to the present invention will now be described in fiirther detail.
[36] Haar-like features and the AdaBoost learning algorithm are two leading lace detection algorithms applied to the present invention. Haar-like features are widely used in lace searching, and numerous prototypes have been trained to accurately represent human laces through the AdaBoost learning algorithm. Thus, it is possible to achieve more efficient lace detection using only the Haar-like features that are trained to represent human laces well through the AdaBoost learning algorithm (reference: Sunghoon Park, Jaeho Lee, Whoiyul Kim, 'Face Recognition Using Haar-like Feature/ LDA', Workshop on Image Processing and Image Understanding).
[37] Haar-like features, which are identifiers proposed by Viola et al., are simple and use simple addition operations. In a Haar-like feature approach, feature values are obtained by summing up the values of pixels in each region of a lace image and weighting and then summing up the regional sums, instead of directly using the values of the pixels of the lace image. Thus, the Haar-like feature approach is usefiil in lace detection based on moving pictures.
[38] FIG. 1 illustrates various prototypes of Haar-like features. Referring to FIG. 1, the locations of Haar-like features may vary in a window according to the characteristics of a lace image. Therefore, the Haar-like features may have various values depending on the type of image to be recognized. Specifically, FIG. 1(1) illustrates prototypes of Haar-like features representing edge characteristics of an image, FIG. 1(2) illustrates prototypes of Haar-like features representing line characteristics of an image, and FIG. 1(3) illustrates prototypes of Haar-like features representing center characteristics of an image.
[39] As described above, in the present invention, the AdaBoost learning algorithm is used for enhancing lace recognition rates. Specifically, the AdaBoost learning algorithm, which is simple but efficient compared to other boost algorithms, is used for more precisely differentiating a lacial area from a non^acial area and boosting the probability of an area of an image determined to be a lacial area. In the AdaBoost learning algorithm, locations of prototypes of Haar-like features can render more detailed lacial characteristics especially at a higher stage of classification. Haar-like features obtained using the AdaBoost learning algorithm are classified in stages, as shown in FIG. 2. Referring to FIG. 2, of a plurality of negative windows 10, those
which are determined as being true are allowed to proceed from node 1 to node 2, from node 2 to node 3, and from node 3 to node 4, and those which are determined as being felse are rejected (20). Only the negative windows 10 which have not been rejected at any of nodes 1 through 4 are additionally processed (30).
[40] The classification of Haar-like features shown in FlG. 2 is carried out in stages in order to realize a more robust lace recognition algorithm. More feature values are generated at a higher stage of classification than at a lower stage of classification. In the present embodiment, a total of 25 stages of classification are performed, and 200 Haar-like features are generated at the 25 stage of classification.
[41] A lace detection method using Haar-like features, according to an exemplary embodiment of the present invention, will now be described more fiilly with reference to FIG. 3.
[42] Referring to FIG. 3, a lace image is received, a window is laid over a lacial area of the lace image, and Haar-like features of the lacial areas are obtained. Thereafter, some of the Haar-like features are selected using the AdaBoost learning algorithm, and the selected Haar-like features are classified and then stored. In a first stage of clas¬ sification, the smallest number of Haar-like features, i.e., 9 Haar-like features, are classified. A total of 25 stages of classification are performed, and a total of 200 Haar- like features are classified in a 25 stage of classification. Data obtained as results of the first through 25 stages of classification is stored as a text file and is used as a hidden cascade in lace recognition. In the present embodiment, the size of the window is set to 24 x 24. In addition, the number of Haar-like features processed at a higher stage of classification is larger than the number of Haar-like features processed at a lower stage of classification, and prototypes of Haar-like features at a higher stage of classification can render more detailed lacial characteristics than at a lower stage of classification (reference: Alexander Kuranov, Rainer Lienhart, and Vadim Pisarevsky, 'An Empirical Analysis of Boosting Algorithms for Rapid Objects with an Extended Set of Haar-like Features', Intel Technical Report MRL-TR-JuIy 02-01, 2002).
[43] The principle of the lace detection method using Haar-like features according to the exemplary embodiment of the present invention will now be described with reference to FIG. 3.
[44] Referring to FIG. 3, an area of an image of one frame is determined as a lacial area using a group of trained lace images 300 while gradually decreasing the size of the image in a pyramid manner. A plurality of candidate areas for the lacial area of the image are generated in the process of recovering the size of the image, and an average
of the candidate areas is output. Facial areas of trained lace images are detected and then differentiated from non^acial areas of the trained lace images using lace classifiers. Thereafter, lace authentication is performed by comparing the detected fecial area with the trained lace images using an HMM.
[45] The HMM algorithm is based on the Markov assumption that the probability of an event at a given moment of time depends only on the given moment of time, rather than any previous moments of time. Markov stages are hidden but may be observable only through other probability processes (reference: F. Samaria and S. Young, 'HMM Based Architecture for Face Identification', ICASSP 98, Vol. 12, pp. 537-543, October 1994).
[46] According to the HMM algorithm, a lace I recognized based on its features.
Prominent features of a lace include the hair, forehead, eyes, nose, and mouth. These fecial features are modeled using a one-dimensional (ID) HMM. Morkov states are dependent on this modeling process and are applied to a lace image in the order from the top to the bottom of the lace image. The lace image is segmented in consideration of such elements of a lace as the eyes, nose, ears, and mouth, whose locations in the lace are fixed. The ID HMM achieves a lace recognition rate of about 85% (references: A. V. Nefian and M. H. Hayes, 'A Hidden Markov Model for Face Recognition', ICASSP 98, Vol. 5, pp. 2721-2724, 1998, and A. V. Nefian and M. H. Hayes, 'Face Detection and Recognition Using Hidden Markov Models', International Conference on Image Processing, 1998)
[47] The ID HMM can be extended to a pseudo two-dimensional (2D) HMM
(reference: S. Kuo and O. Agazzi, 'Keyword Spotting in Poorly Printed Documents Using Pseudo 2D Hidden Markov Models', IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 16, pp. 842-848, August 1994). The pseudo 2D HMM is generated through the expansion of each block of a lace image. Referring to FIG. 4, a superstate 42 is divided into a number of states 44. Thereafter, the probabilities of the states 44 are calculated, and then the probability of the superstate 42 is calculated based on the probabilities of the states 44. In this manner, subsequent superstates are processed. FIG. 4 illustrates the calculating of the probabilities of superstates of a lace image in the order of the forehead, eyes, nose, mouth and chin of the lace image.
[48] FIG. 5 is a block diagram illustrating a process of training lace data for lace detection and authentication and a process of generating a database of lace data in the lace detection and authentication method according to the exemplary embodiment of the present invention. Referring to FIG. 5, a block extraction unit 510 receives a
training image 500 of an N-th person and extracts a block from the received trained
N N N image. A feature extraction unit 520 extracts features O , O , ..., O from the
1 2 J extracted block received from the block extraction unit 510 and outputs them to an initial parameter evaluation unit 54) and a Baum- Welch parameter reevaluation unit 550. The initial parameter evaluation unit 54) receives an HMM model from a
N N N prototype HMM modeling unit 530, evaluates the extracted features O , O , ..., O
1 2 J received from the feature extraction unit 520 using the HMM model λ , and outputs o the evaluation results λ to the Baum- Welch parameter reevaluation unit 550. The mlt N
Baum- Welch parameter reevaluation unit 550 outputs a trained lace image λ based on
N N N the extracted features O , O , ..., O received from the feature extraction unit 520
1 2 J and the evaluation results 1 received from the initial parameter evaluation unit 54).
N lmt
The trained lace image λ output from the Baum- Welch parameter reevaluation unit 550 is stored in a trained lace image database (not shown).
[49] FlG. 6 is a block diagram of a lace detection and authentication apparatus according to an exemplary embodiment of the present invention. Referring to FlG. 6, føcial areas are automatically extracted from moving pictures using Haar-like features and then are stored as BMP files. Six lace images of each person to be authenticated can be additionally registered with a database of previously-gathered lace images. Referring to FlG. 6, a lace image 62 is automatically extracted from a moving picture 6D of a person to be authenticated and then stored. Parameters are extracted from the extracted lace image 62 by passing the extracted lace image 62 through a block extraction unit 64 and a feature extraction unit 66. The extracted parameters are compared with parameters stored in the database of lace images by approximation model calculation units 71, ..., 12. Thereafter, one of the lace images stored in the database that is most similar to the extracted lace image 62 is selected by a maximum probability selection unit 80 and then output via an output unit 90.
[50] Table 1 shows trained location information and critical values of prototypes of
Haar-like features in a 24 x 24 window, and Table 2 shows the results of testing the lace detection and authentication apparatus according to the exemplary embodiment of the present invention.
[51] Table 1 extern const char* FaceCascade[]={ /*Stage 0*/
[52] Table 2
[53] FIG. 7 is a diagram illustrating lace authentication results obtained using the lace detection and authentication method according to an exemplary embodiment of the present invention. Referring to FIG. 7, 48 lace images of 8 people (6 lace images per
person) and a trained database of 24) lace images of 4) people (6 lace images per person) were used in experiments for testing the lace detection and authentication apparatus according to the exemplary embodiment of the present invention. The size of the forehead in lace images of people who ήiled to be authenticated is larger than in lace images of people who were successfiilly verified, as shown in FlG. 7. Because of a larger forehead, the eyes in the lace images of those who føiled to be authenticated deviate from expected locations when segmenting the corresponding lace images in a 2D HMM-based training process.
[54] Examples of the lace detection and authentication method according to the exemplary embodiment of the present invention will now be described more fiilly with reference to FIGS. 8 through 11. In FIGS. 8 through 11, like reference numerals represent like elements.
[55] FIG. 8 is a diagram illustrating an example of the lace detection and authentication method according to the exemplary embodiment of the present invention that uses a wearable personal station (WPS) 80, which is a personal portable terminal. Referring to FIG. 8, the WPS 80 captures a lace image from a moving picture, wirelessly transmits the captured lace image to a server 90 using a JPEG codec. The server 90 stores a database of lace images, lace detection algorithms, and lace authentication algorithms. In addition, the server 90 includes a lace authentication unit, which compares the captured lace image transmitted from the WPS 80 with the lace images stored in the server and authenticates the captured lace image based on the comparison results. Specifically, the WPS 80 receives a lace image from a camera 82 installed therein, processes the lace image, and wirelessly transmits the processed lace image to the server 90 via Bluetooth or wireless Internet (86). The server 90 trains and classifies a number of lace images in 25 stages using Haar-like features and the AdaBoost learning algorithm, which are lace detection algorithms stored in the server 90, so that a more detailed lacial area mask can be generated at a higher stage of classification than at a lower stage of classification. Thereafter, the server 90 receives an image to be processed, determines a portion of the image to be processed that matches with the lacial area mask as a lacial area while gradually reducing the size of the image to be processed. Thereafter, the server 90 performs lace authentication (92) by comparing the portion of the image to be processed determined as a lacial area with the lace images stored in the server 90 using an HMM algorithm, which is the lace au¬ thentication algorithm stored in the server 90.
[56] FIG. 9 is a diagram illustrating another example of the lace detection and au-
thentication method according to the exemplary embodiment of the present invention that uses a WPS 80. Referring to HG. 9, the WPS 83 automatically extracts &ce data from an image to be authenticated using a lace detection algorithm stored therein and wirelessly transmits the extracted lace data using a JPEG codec to a server 90. The server 90 stores a database of lace images and a lace authentication algorithm. In addition, the server 90 includes a lace authentication unit, which authenticates the image to be authenticated by comparing the extracted lace data with lace data stored therein. Specifically, the WPS 80 receives a lace image from a camera 82 installed therein, extracts lace data from the lace image using Haar-like features, and transmits (86) the extracted lace data to the server 90 via Bluetooth or wireless Internet. In addition, the WPS 80 trains and classifies a number of lace images in 25 stages using Haar-like features and the AdaBoost learning algorithm, which are lace detection algorithms stored in the WPS 80, so that a more detailed lacial area mask can be generated at a higher stage of classification than at a lower stage of classification. Thereafter, the WPS 80 receives an image to be processed, determines a portion of the image to be processed that matches with the lacial area mask as a lacial area while gradually reducing the size of the image to be processed. Thereafter, the WPS 80 performs lace authentication (92) by comparing the portion of the image to be processed determined as a lacial area with the lace images stored in the server 90 using an HMM algorithm, which is the lace authentication algorithm stored in the server 90. [57] FIG. 10 is a diagram illustrating another example of the lace detection and au¬ thentication method according to the exemplary embodiment of the present invention that uses a WPS 80. Referring to FIG. 10, the WPS 80 includes a &ce detection unit, which detects a lace image and a storage unit, which stores a number of lace images and lace detection algorithms. A server 90 stores a database of lace images and lace authentication algorithms and includes a lace authentication unit, which authenticates the detected lace image by comparing the detected lace image with the lace images stored in the server 90. The WPS 80 receives a lace image from a camera 82 installed therein, detects lace data from the lace image using Haar-like features, and outputs parameters of features of the detected lace data (84). The detected lace data is transmitted to the server 90 via Bluetooth or wireless Internet. As described above, the WPS 80 trains and classifies a number of lace images in 25 stages using Haar-like features and the AdaBoost learning algorithm, which are lace detection algorithms stored in the WPS 80, so that a more detailed lacial area mask can be generated at a higher stage of classification than at a lower stage of classification. Thereafter, the
WPS 80 receives an image to be processed, determines a portion of the image to be processed that matches with the lacial area mask as a lacial area while gradually reducing the size of the image to be processed. Thereafter, the server 90 performs lace authentication (86) by comparing parameters of the portion of the image to be processed determined as a lacial area with parameters of the lace images stored in the database using an HMM algorithm, which is a lace authentication algorithm stored in the server 90.
[58] FlG. 11 is a diagram illustrating another example of the lace detection and au¬ thentication method according to the exemplary embodiment of the present invention that uses a WPS 80. Referring to FlG. 10, the WPS 80 includes a &ce detection unit, which detects a lace image, a storage unit, which stores a number of lace images and lace detection and authentication algorithms, and a lace authentication unit, which au¬ thenticates the detected lace image by comparing the detected lace image with the lace images stored therein. The WPS 80 receives a lace image from a camera 82 installed therein, detects lace data from the lace image using Haar-like features, and outputs parameters of features of the detected lace data (84). As described above, the WPS 80 trains and classifies a number of lace images in 25 stages using Haar-like features and the AdaBoost learning algorithm, which are lace detection algorithms stored in the WPS 80, so that a more detailed lacial area mask can be generated at a higher stage of classification than at a lower stage of classification. Thereafter, the WPS 80 receives an image to be processed, determines a portion of the image to be processed that matches with the lacial area mask as a lacial area while gradually reducing the size of the image to be processed. Thereafter, the WPS 80 performs lace authentication (86) by comparing parameters of the portion of the image to be processed determined as a lacial area with parameters of the lace images stored in the database using an HMM algorithm, which is a lace authentication algorithm stored in the WPS 80.
[59] The structure, operation, and performance of a lace recognition system using Haar- like features and an HMM algorithm have been described above. The lace recognition system is expected to be very usefiil because it can perform lace recognition processes not only on still images but also on moving pictures even in real time. The lace recognition system can verify laces very quickly using Haar-like features and can be easily applied to moving pictures because of its short lace verification time. In ad dition, since the lace recognition system applies lace area data extracted from a lace image to an HMM, it is possible to quickly perform lace recognition on video image data in real time. Moreover, the lace recognition system can be realized based on com-
munications between a portable device and a server. Face recognition rates of the lace recognition system can be boosted by generating HMMs using more detailed in¬ formation on the location of the eyes and the mouth and the skin color of laces to be verified.
Industrial Applicability [(D] As described above, according to the present invention, a terminal automatically extracts a lacial area from an image stored therein and wirelessly transmits parameters of features of the extracted lacial area to a server. Then, the server decides whether to authenticate the extracted lacial area based on a result of comparing the extracted lacial area with a database of lace images stored therein. The terminal can extract lace data and lacial features from a moving picture, and the server can store a considerable amount of data. Thus, it is possible to provide efficient solutions for identifying and authenticating people, providing them with a log-on fiinction, and holding video conferences.
Claims
[1] A lace detection and authentication apparatus comprising: a personal portable terminal, which processes a lace image of a moving picture input thereto, stores the processed lace image, and wirelessly transmits the processed lace image; and a server, which comprises a storage unit storing a lace database, lace detection algorithms, and lace authentication algorithms and a lace authentication unit au¬ thenticating lace data extracted from the lace image that is wirelessly transmitted by the personal portable terminal by comparing the extracted lace data with lace data stored in the lace database and then outputting the authentication results.
[2] A lace detection and authentication apparatus comprising: a personal portable terminal, which comprises a storage unit storing lace detection algorithms and a lace detection unit detecting lace data from a moving picture input thereto; and a server, which comprises a storage unit storing a lace database and lace au¬ thentication algorithms and a lace authentication unit authenticating the detected lace data by comparing the detected lace data with lace data stored in the lace database and then outputting the authentication results.
[3] A lace detection and authentication apparatus comprising: a personal portable terminal, which comprises a storage unit storing lace data and lace detection algorithms and a lace detection unit detecting lace data from a moving picture input thereto and wirelessly transmitting parameters of features of the detected lace data; and a server, which comprises a storage unit storing a lace database and lace au¬ thentication algorithms and a lace authentication unit authenticating the detected lace data based on the parameters of the features that is wirelessly transmitted by the personal portable terminal by comparing the detected lace data with lace data stored in the lace database and then outputting the authentication results.
[4] A lace detection and authentication apparatus comprising: a storage unit, which stores lace data in a personal portable terminal; a lace detection unit, which detects lace data from a moving picture input to the personal portable terminal and outputs parameters of features of the detected lace data; and a lace authentication unit, which authenticates the detected lace data based on
the parameters of the features by comparing the detected lace data with lace data stored in a lace database and then outputs the authentication results.
[5] The lace detection and authentication apparatus of any of claims 1 through 4, wherein lace data is detected from a moving picture by laying a window over a lacial area of the moving picture, obtaining Haar-like feature values of the lacial area, and then classifying and storing the Haar-like feature values using prototypes of Haar-like features and the AdaBoost learning algorithm.
[6] The lace detection and authentication apparatus of any of claims 1 through 4, wherein a lace image is authenticated by generating a hidden Markov model (HMM) based on features of lace data detected from the lace image using an HMM algorithm and then sequentially applying the HMM model to the lace image in the order from the top to the bottom of the lace image.
[7] The lace detection and authentication apparatus of claim 6, wherein in the HMM algorithm, a two dimensional (2D) HMM is generated through the expansion of each block of the lace image by dividing each superstate into a plurality of states, calculating the probabilities of the states, and calculating the probability of a superstate based on the probabilities of the states into which the superstate is divided.
[8] A lace detection and authentication method comprising: storing a lace database, lace detection algorithms, and lace authentication algorithms in a server; processing a lace image of a moving picture input to a personal portable terminal, storing the processed lace image, and wirelessly transmitting the processed lace image from the personal portable terminal to the server; and authenticating lace data extracted from the lace image that is wirelessly transmitted by the personal portable terminal by comparing the extracted lace data with lace data stored in the lace database stored in the server and then outputting the authentication results.
[9] A lace detection and authentication method comprising: storing a lace database and lace authentication algorithms in a server; detecting lace data from a moving picture input to a personal portable terminal using lace detection algorithms stored in the personal portable terminal and wirelessly transmitting the detected lace data from the personal portable terminal to the server; and authenticating the detected lace data that is wirelessly transmitted by the
personal portable terminal by comparing the detected lace data with lace data stored in the lace database stored in the server and then outputting the au¬ thentication results.
[10] A lace detection and authentication method comprising: storing a lace database and lace authentication algorithms in a server; detecting lace data from a moving picture input to a personal portable terminal using lace detection algorithms stored in the personal portable terminal and wirelessly transmitting parameters of features of the detected lace data from the personal portable terminal to the server; and authenticating the detected lace data based on the parameters of the features that is wirelessly transmitted by the personal portable terminal by comparing the detected lace data with lace data stored in the lace database stored in the server and then outputting the authentication results.
[11] A lace detection and authentication method comprising : storing lace data in a personal portable terminal; detecting lace data from a moving picture input to the personal portable terminal and outputting parameters of features of the detected lace data; and authenticating the detected lace data based on the parameters of the features by comparing the detected lace data with lace data stored in a lace database and then outputting the authentication results.
[12] The lace detection and authentication method of any of claims 8 through 11, wherein lace data is detected from a moving picture by laying a window over a lacial area of the moving picture, obtaining Haar-like feature values of the lacial area, and then classifying and storing the Haar-like feature values using prototypes of Haar-like features and the AdaBoost learning algorithm.
[13] The lace detection and authentication method of any of claims 8 through 11, wherein a lace image is authenticated by generating a hidden Markov model (HMM) based on features of lace data detected from the lace image using an HMM algorithm and then sequentially applying the HMM model to the lace image in the order from the top to the bottom of the lace image.
[14] The lace detection and authentication method of claim 13, wherein in the HMM algorithm, a two dimensional (2D) HMM is generated through the expansion of each block of the lace image by dividing each superstate into a plurality of states, calculating the probabilities of the states, and calculating the probability of a superstate based on the probabilities of the states into which the state is divided.
Applications Claiming Priority (4)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| KR1020040096172A KR100702226B1 (en) | 2004-11-23 | 2004-11-23 | Apparatus and method for detecting and authenticating facial data |
| KR10-2004-0096171 | 2004-11-23 | ||
| KR1020040096171A KR100702225B1 (en) | 2004-11-23 | 2004-11-23 | Apparatus and method for detecting and authenticating facial data |
| KR10-2004-0096172 | 2004-11-23 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2006057475A1 true WO2006057475A1 (en) | 2006-06-01 |
Family
ID=36498191
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/KR2004/003480 WO2006057475A1 (en) | 2004-11-23 | 2004-12-28 | Face detection and authentication apparatus and method |
Country Status (1)
| Country | Link |
|---|---|
| WO (1) | WO2006057475A1 (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8643691B2 (en) | 2008-05-12 | 2014-02-04 | Microsoft Corporation | Gaze accurate video conferencing |
| CN104298969A (en) * | 2014-09-25 | 2015-01-21 | 电子科技大学 | Crowd scale statistical method based on color and HAAR feature fusion |
| CN104978550A (en) * | 2014-04-08 | 2015-10-14 | 上海骏聿数码科技有限公司 | Face recognition method and system based on large-scale face database |
| CN105260744A (en) * | 2015-10-08 | 2016-01-20 | 北京航空航天大学 | Automatic on-line diagnosis method for freight train coupler tail cotter position faults and system |
| CN106778445A (en) * | 2015-11-20 | 2017-05-31 | 沈阳新松机器人自动化股份有限公司 | Service robot vision based on Face datection leads method |
| CN106971193A (en) * | 2016-11-23 | 2017-07-21 | 南京理工大学 | Object detection method based on structural type Haar and Adaboost |
| CN107315993A (en) * | 2017-05-10 | 2017-11-03 | 苏州天平先进数字科技有限公司 | A kind of peephole system and its face identification method based on recognition of face |
| CN109711403A (en) * | 2018-12-18 | 2019-05-03 | 电子科技大学 | Mould detection method in a kind of leukorrhea based on Haar feature and Adaboost |
| US10678903B2 (en) | 2016-05-02 | 2020-06-09 | Hewlett-Packard Development Company, L.P. | Authentication using sequence of images |
| US11875268B2 (en) | 2014-09-22 | 2024-01-16 | Samsung Electronics Co., Ltd. | Object recognition with reduced neural network weight precision |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6088451A (en) * | 1996-06-28 | 2000-07-11 | Mci Communications Corporation | Security system and method for network element access |
| US20030142854A1 (en) * | 2002-01-30 | 2003-07-31 | Samsung Electronics Co., Ltd. | Apparatus and method for providing security in a base or mobile station by using detection of face information |
| US6633655B1 (en) * | 1998-09-05 | 2003-10-14 | Sharp Kabushiki Kaisha | Method of and apparatus for detecting a human face and observer tracking display |
-
2004
- 2004-12-28 WO PCT/KR2004/003480 patent/WO2006057475A1/en active Application Filing
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6088451A (en) * | 1996-06-28 | 2000-07-11 | Mci Communications Corporation | Security system and method for network element access |
| US6633655B1 (en) * | 1998-09-05 | 2003-10-14 | Sharp Kabushiki Kaisha | Method of and apparatus for detecting a human face and observer tracking display |
| US20030142854A1 (en) * | 2002-01-30 | 2003-07-31 | Samsung Electronics Co., Ltd. | Apparatus and method for providing security in a base or mobile station by using detection of face information |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8643691B2 (en) | 2008-05-12 | 2014-02-04 | Microsoft Corporation | Gaze accurate video conferencing |
| CN104978550A (en) * | 2014-04-08 | 2015-10-14 | 上海骏聿数码科技有限公司 | Face recognition method and system based on large-scale face database |
| CN104978550B (en) * | 2014-04-08 | 2018-09-18 | 上海骏聿数码科技有限公司 | Face identification method based on extensive face database and system |
| US11875268B2 (en) | 2014-09-22 | 2024-01-16 | Samsung Electronics Co., Ltd. | Object recognition with reduced neural network weight precision |
| CN104298969A (en) * | 2014-09-25 | 2015-01-21 | 电子科技大学 | Crowd scale statistical method based on color and HAAR feature fusion |
| CN105260744A (en) * | 2015-10-08 | 2016-01-20 | 北京航空航天大学 | Automatic on-line diagnosis method for freight train coupler tail cotter position faults and system |
| CN105260744B (en) * | 2015-10-08 | 2018-08-14 | 北京航空航天大学 | The automatic on-line diagnostic method and system of a kind of goods train coupler yoke key position failure |
| CN106778445A (en) * | 2015-11-20 | 2017-05-31 | 沈阳新松机器人自动化股份有限公司 | Service robot vision based on Face datection leads method |
| US10678903B2 (en) | 2016-05-02 | 2020-06-09 | Hewlett-Packard Development Company, L.P. | Authentication using sequence of images |
| CN106971193A (en) * | 2016-11-23 | 2017-07-21 | 南京理工大学 | Object detection method based on structural type Haar and Adaboost |
| CN107315993A (en) * | 2017-05-10 | 2017-11-03 | 苏州天平先进数字科技有限公司 | A kind of peephole system and its face identification method based on recognition of face |
| CN109711403A (en) * | 2018-12-18 | 2019-05-03 | 电子科技大学 | Mould detection method in a kind of leukorrhea based on Haar feature and Adaboost |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Mahmood et al. | WHITE STAG model: Wise human interaction tracking and estimation (WHITE) using spatio-temporal and angular-geometric (STAG) descriptors | |
| Khan et al. | Facial recognition using convolutional neural networks and implementation on smart glasses | |
| Rodriguez et al. | Face authentication using adapted local binary pattern histograms | |
| Kollreider et al. | Real-time face detection and motion analysis with application in “liveness” assessment | |
| US8837786B2 (en) | Face recognition apparatus and method | |
| CN109800643B (en) | Identity recognition method for living human face in multiple angles | |
| Mir et al. | Biometrics verification: a literature survey | |
| JP2004133889A (en) | Method and system for recognizing image object | |
| El Khiyari et al. | Age invariant face recognition using convolutional neural networks and set distances | |
| Meenakshi | Real-Time Facial Recognition System—Design, Implementation and Validation | |
| Kim et al. | Teeth recognition based on multiple attempts in mobile device | |
| WO2006057475A1 (en) | Face detection and authentication apparatus and method | |
| Stylianou et al. | GMM-based multimodal biometric verification | |
| Naik | Detailed survey of different face recognition approaches | |
| Islam et al. | A fully automatic approach for human recognition from profile images using 2D and 3D ear data | |
| KR100702226B1 (en) | Apparatus and method for detecting and authenticating facial data | |
| Suneetha | A survey on video-based face recognition approaches | |
| KR20090005920A (en) | Method and apparatus for generating an object descriptor using curvature gabor filter | |
| Singha et al. | Study and analysis on biometrics and face recognition methods | |
| Olufade et al. | Biometric authentication with face recognition using principal component analysis and feature based technique | |
| Park et al. | Multimodal priority verification of face and speech using momentum back-propagation neural network | |
| Shatnawi et al. | The impact of the number of eigen-faces on the face recognition accuracy using different distance measures | |
| Hong et al. | Real time face detection and recognition system using Haar-like feature/HMM in ubiquitous network environments | |
| Park et al. | Multi-modal human verification using face and speech | |
| KR100702225B1 (en) | Apparatus and method for detecting and authenticating facial data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| AK | Designated states |
Kind code of ref document: A1 Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BW BY BZ CA CH CN CO CR CU CZ DE DK DM DZ EC EE EG ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NA NI NO NZ OM PG PH PL PT RO RU SC SD SE SG SK SL SM SY TJ TM TN TR TT TZ UA UG US UZ VC VN YU ZA ZM ZW |
|
| AL | Designated countries for regional patents |
Kind code of ref document: A1 Designated state(s): GM KE LS MW MZ NA SD SL SZ TZ UG ZM ZW AM AZ BY KG KZ MD RU TJ TM AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HU IE IS IT LT LU MC NL PL PT RO SE SI SK TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG |
|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application | ||
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| 32PN | Ep: public notification in the ep bulletin as address of the adressee cannot be established |
Free format text: NOTING OF LOSS OF RIGHTS PURSUANT RULE 69(1) EPC, EPO FORM 1205A DATED 16.08.07 |
|
| 122 | Ep: pct application non-entry in european phase |
Ref document number: 04808609 Country of ref document: EP Kind code of ref document: A1 |