This application claims priority based on a Japanese patent application, No. 209-138778 filed on Jun. 10, 2009, the entire contents of which are incorporated herein by reference.
The subject discussed herein relates to a biometrics authentication method for authenticating an individual by using personal biometric information.
In an individual authentication system using biometric information, the biometric information of an individual is acquired at the time of initial registration and information called feature is extracted and registered. The feature registered is termed a template. During authentication, the biometric information is again captured from the individual to extract the feature which in turn is collated with the precedently registered template to confirm whether or not the individual is the person in question. When, in a system having a client unit (simply referred to as a client) and a server unit (simply referred to as a server) coupled together through a network, the server carries out biometrics authentication of a user being present on the client side, the server holds a template typically. The client acquires biometric information the user, extracts a feature to transmit it to the server and the server collates the feature with the template to confirm whether or not the user is the person in question.
The template is however, effective information to identify an individual and is thereof required to be managed stringently as personal information, incurring high costs of management.
Further, even if the information is placed in stringent management, many persons having their privacy in mind will psychologically be reluctant to resister the template. In addition, the biometric information remains unchanged through life and cannot be changed easily in contrast to the password and the encrypted key. Conceivably, the biometric information may be exchanged for different one but the number of pieces of biometric information of one kind a single individual has is limited (for example, the number of fingerprints is limited to the number of fingers), failing to provide essential solving measures. Accordingly, in case the template is leaked facing a risk of forgery, their arises a problem that the biometrics authentication cannot be used safely ever since. Furthermore, if the same biometric information is registered in a different system, even the different system will be exposed to the menace.
Under the circumstances, a method as described in “Enhancing security and privacy in biometric-based authentication systems by N. K. Ratha, J. H. Connell and R. M. Bolle, IBM system Journal 40(3), 2001 (reference 1) has been proposed, according to which during registration of biometric information, a feature is transformed with the help of a particular function (a kind of encryption) and a secret parameter (a kind of encrypted key) the client has and the original information is taken in custody as a template by the server while being kept to be concealed and during authentication, a feature of biometric information newly extracted by the client is transformed with the same function and parameter as those above and transmitted to the server, and the server collates the received feature with the template while keeping them transformed (termed cancelable biometrics authentication).
According to the above method, the client keeps the transform parameter secret and so the server cannot know the original feature even during authentication and the privacy of individual can be protected.
Further, even if the template is leaked, the security can be assured by changing the transform parameter, preparing a template again and registering it. In addition, when the same biometric information is used for a different system, templates transformed with the help of different parameters, respectively, are registered so that even with one template leaked, the security of the different system can be prevented from being degraded.
According to the aforementioned reference 1, the client transforms a feature image extracted from biometric information of a user with the help of a transform parameter P to prepare a transformed feature T and registers it in the server. The transform parameter P is saved in a Smart card, for example, to enable the user to manage it in secret.
During authentication, a feature image y the client newly extracts from the user's biometric information is transformed with the transform parameter P read out the user's Smart card to prepare a transformed feature V which in turn is transmitted the server. The server calculates the degree of similarity (hereinafter simply referred to as similarity) between the T and the V to decide match/mismatch.
In this manner, by transmitting to the server the T and V obtained by transforming the x and y with the secret transform parameter P, the server can be allowed to execute the matching process while x and y are concealed from the server.
Incidentally, when applying the cancelable biometrics authentication a system in which biometric information is inputted (without inputting any user ID) and a user ID is outputted (hereinafter referred to as 1:N authentication), matching with all transformed features T's in a DB of the server simply needs to be executed. But in the case of a large-scale DB, for matching with all T's, time to process the matching with all of the T's is prolonged in general, making the system unpractical disadvantageously.
To cope with this problem, a biometrics authentication technology is disclosed in which when applying cancelable biometrics authentication to an 1:N authentication system, the time required for the server to execute the 1:N matching can be shortened to speed up the process.
As an example is disclosed, in a method of registering biometric information based on a biometrics authentication method for authenticating an individual without using an ID of the individual but with the use of biometric information in a system including a client and a server, dummy features respectively representing groups each constituted by similar features are set in advance, the client performs a rough process for specifying, on the basis of a dummy feature, a group to which a feature for registration should belong and the server performs a process for registering a transformed feature or registration on the specified group.
As another example is disclosed, in a method of authenticating biometric information based on a biometrics authentication method for authenticating an individual without using an ID of the individual but with the use of biometric information in a system including a client and a server, dummy features respectively representing groups each constituted by similar features are set in advance, the client performs an rough process for specifying, on the basis of a dummy feature, a group to which a feature for authentication should belong and the server performs a detailed process for specifying any one of similar features in respect of the specified group.
Further, the feature is a transformed feature obtained by transforming biometric information extracted from an individual with the help of a predetermined parameter.
As a more specified example of method for biometrics authentication disclosed, in a method for biometrics authentication in which a client terminal transforms a feature for authentication extracted from biometric information of an individual with a transform parameter obtain a transformed feature for authentication and transmits it to an authentication server and the authentication server includes a database adapted to store a plurality of transformed features for registration and information specifying individuals by making the correspondence between them and searches the database to specify an individual corresponding to a transformed feature for registration which is the most similar to the transformed feature for authentication,
the client terminal stores precedently in the database on the client side dummy features on the basis of which the features for registration and the features for authentication of the biometric information are classified into any of a plurality of groups, an index for identifying a group and transform parameters corresponding to the respective dummy features;
the authentication server includes a database of transformed features for registration adapted to store the transformed features for registration transformed from the features for registration with the help of transform parameters and information for specifying the individual from which extraction of the features for registration originates while classifying them into any of the plural groups; and
for the process during registration of the biometric information, the client terminal comprises the steps of:
extracting a feature for registration from biometric information acquired from an individual;
searching the database on the client side specify one dummy feature for registration similar to the feature for registration;
transforming the feature for registration by using a transform parameter corresponding to the specified dummy feature for registration to prepare a transformed feature for registration; and
transmitting to the authentication server the transformed feature for registration, an index for registration adapted to identify the specified group of dummy features for registration and information for specifying the individual from which the extraction of features for registration originates, and
the authentication server comprises the step of:
registering, at a group corresponding to the index for registration in the database for transformed features for registration, the received transformed feature for registration and the information specifying the individual while making the correspondence between them.
As furthermore specified example of method for biometrics authentication disclosed, the client terminal includes the steps of:
extracting a feature for authentication from biometric information acquired from a individual;
searching a database on the client side to specify one dummy feature for authentication similar to the feature for authentication;
transforming the feature for authentication by using a transform parameter corresponding to the specified dummy feature for authentication to prepare a transformed feature for authentication; and
transmitting to the authentication server the transformed feature for authentication and the index for authentication adapted to identify a group of specified dummy features for authentication, and
the authentication server includes the steps of:
searching a group corresponding to the index for authentication in the database for transformed feature for registration to specify a transformed feature for registration which is the most similar the transformed feature for authentication;
specifying the individual having the correspondence with the specified transformed feature for registration; and
transmitting to the client terminal the information concerning the specified individual.
According to the above examples, in the 1:N biometrics authentication system inputting the biometric information and outputting the user ID, time required for 1:N matching on the server side can be shortened and speedup of the processing can be achieved.
According to the examples of disclosure, a 1:N biometrics authentication system capable of executing high-speed authentication while keeping the user's biometric information concealed can be realized.
These and other benefits are described throughout the present specification. A further understanding of the nature and advantages of the realized by reference to the remaining portions of the specification and the attached drawings.
An embodiment of the biometrics authentication system will now be described with reference to the accompanying drawings.
The present embodiment will be described by way of example of a 1:N cancelable biometrics authentication system in which a cancelable biometrics authentication is applied to a 1:N authentication.
Referring first to
In the present embodiment, the 1:N cancelable biometrics authentication system comprises a client terminal (hereinafter simply referred to as a client) 100 for performing acquisition of biometric information, extraction of feature, similarity search to a dummy feature and transformation of the features during registration/authenticating, an auxiliary data DB preparation terminal 110 for preparing a DB of auxiliary data inclusive of the dummy feature and an authentication server (hereinafter simply referred to as a server) 120 for storing templates and performing 1:N matching. These components are coupled to one another through a network such as Internet or Intranet.
The client 100 includes sensor 107 managed by a user on his or her own or by a reliable third party and adapted to acquire biometric information (for example, fingerprints or veins). Taking credit settlement in a retail shop, for instance, the client 100 may structurally be a credit terminal managed by the shop and the server 120 may be a server machine managed by a credit service company.
Then, the client 100 includes a feature extraction unit 101 which extracts a feature from the biometric information captured by the sensor 101, an auxiliary data DB 103 in which plurality of dummy features are stored in advance, a similarity search unit 102 which performs similarity search to dummy features in the auxiliary data DB 103, a transform parameter DB 105 which stores transform parameters corresponding to respective dummy features, a feature transform unit 104 which transforms a feature for registration or authentication so as to prepare a transformed feature for registration (hereinafter referred to as a template) or a transformed feature for authentication, and a communication unit 106 which communicates with the server.
The biometric information referred to herein is data indicative of, for example, a fingerprint image, a vein image and an iris image and the feature includes an image of, for example, a fingerprint or vein which undergoes an emphasis process for binary digitization or it includes a bit string called an iris code prepared from the iris image. Available as the similarity between the two features is that calculated through, for example, cross-correlation. Then, known as the matching algorithm for calculation of similarity on the basis the cross-correlation is an algorithm which applies a special transformation to the two features in order to calculate a similarity while keeping the original features concealed and keeping them unrecoverable to their original forms (correlation invariant random filtering). Details of the correlation invariant random filtering are disclosed in JP-A-2007-293807 reference 2) and “Vulnerability Analysis and Improvement of Cancelable Biometrics for Image Matching” by Hirata et al, SCIS2007 Preliminary Papers CD-ROM (reference 3).
The auxiliary data DB preparation terminal 110 includes a feature DB 112 representing a DB of features from which preparation of dummy features originates, a clustering unit 111 which applies clustering to a set of features in the feature DB, a dummy feature preparation unit 113 which prepares dummy features by making use of the results of clustering, an auxiliary data preparation unit 115 which prepares auxiliary data used for performing similarity search to the dummy feature in the client, an auxiliary data DB 114 which stores auxiliary data and a communication unit 116 which communicates with the client 100.
The server 120 includes a communication unit 121 which communicates with the client 100, database 122 for management of templates (template DB. The database will hereinafter be abbreviated as DB), a registration unit 123 which registers in the template DB 122 a template received from the client, and a 1:N matching unit 124 which collates a transformed feature received from the client with a plurality of templates in the template DB 122.
Illustrated in
The functional construction of the client 100, auxiliary data DB preparation terminal 110 and server 120 and individual processes to be described later as well can be materialized when each corresponding CPU 600 executes respective programs stored in the memory 601 and the HDD 602. The respective programs may be stored in the corresponding memory 601 and HDD 602 or may be introduced, as necessary, from a different unit by way of a mountable/dismountable memory medium utilizable by the unit, a communication network representing a communication medium or a carrier wave or a digital signal propagating on the communication network.
Next, by making reference to
In the phase of system development, biometric information is collected in advance from an interested person such as a developer to prepare a set {x[i]} of features which in turn is stored in the feature DB 112. For example, if the biometric information is a finger vein image, a method for feature extraction described in “Feature extraction of finger-vein patterns based on repeated line tracking and its application to personal identification” by N. Miura, A. Nagasaka and T. Miyatake, Machine Vision and Applications 15(4) (2004), 194-203 (reference 5) can be used.
The clustering unit 111 reads the set {x[i]} of features to execute clustering (S201). Through clustering, features similar to one another are grouped. For details of the clustering, one may refer to reference 4. Where a set of groups of features is represented by {G[j]}, features x[j, k] are included in the G[j]. In these expressions, i an index for a feature in the feature DB 112, j is an index for the group of clustered/resembling features and k is an index for a feature contained in the group j.
As a specific method for clustering, a k-means method, for example, can be used. The k-means method will be outlined here. The k-means method is a typical process for non-hierarchal cluster analysis and typically it presupposes the Euclidean space. In the method, the number of clusters is designated in advance (here k) and objects are divided into k classes. The non-similarity is defined by the square of a Euclidean distance and the criteria of classification is the non-similarity between the center of a cluster and each object. The k-means method can designate the number of clusters and hence it is meritorious in that a bias in the number of objects among individual clusters can be mitigated. Algorithm of the k-means method is as follows:
in step 1 [for initial value], centers or initial divisions of k clusters are given at random;
in step 2 [for allotment], each object is allotted to the closest cluster center; and
in step 3 [for update of center], the program ends if allotments of all objects remain unchanged from those in the one preceding step. If not so, the centroid of each cluster is exchanged for a new center and the program returns to step 2.
In the program, respective coordinates at the center of a cluster are represented by a weighted means of coordinate values of objects contained in the cluster and respective coordinate values of the centroid in the cluster are represented by an average devoid of weighing.
The dummy feature preparation unit 113 prepares centroids D[j] in respect of the individual groups G[j] (S 202). The centroid signifies a feature indicative of the barycenter of features contained in the group. For more information, reference is to be made to reference 4. A set of centroids {D[j]} thus prepared is written as a set of dummy features {D[i]} into the auxiliary data DB 114 (S203).
Next, a method for preparation of the transform parameter DB 105 in the present embodiment will be described. Here, transform parameters are individually prepared and made to correspond to the individual dummy features D[i]. A transform parameter corresponding to a dummy feature D[i] is represented by an R[i]. As for a method for preparation of the transform parameter R[i], the method described in reference 2 or 3 can be used in the case of a finger vein image, for instance. Specifically, it is sufficient that a two-dimensional random filter in which values of individual pixels are random numbers is prepared. The random numbers can be generated using a pseudo-random number generator, for example. This random filter is used to provide the transform parameters R[i].
As an example, Locality-Sensitive Hashing (hereinafter termed LSH) described in “Approximate nearest neighbors towards removing the curse of dimensionality” by P. Indy k and R. Motwani, Proc. of the Symposium on Theory of Computing, 1988 (reference 6) is used for the method of performing similarity search to dummy feature in the present embodiment. Another method may be used for similarity search. For example, “Approximating and Eliminating Search Algorithm” (hereinafter termed AESA) described in “An algorithm for finding nearest neighbours in (approximately) constant average time” by Vidal Ruiz, Pattern Recognition Letters, pp. 145-157, 1986 (reference 7) may be available.
In the LSH, a hash table is necessary. By making reference to
The auxiliary data preparation unit 115 reads a dummy feature set {D[i]} stored in the auxiliary DB 114 (S301).
The auxiliary data or preparation unit 115 prepares a hash table (S302). Here, a method described in “Locality-sensitive hashing scheme based on p-stable distributions” by M. Datar, N. Immorlica, P. Indyk and V. Mirrokni, Proc. of the ACM Symposium on Computational Geometry, 2004 (reference 8) is used exemplarily. Another method may be available. In reference 8, the hash is prepared pursuant to the following equation in relation to a feature vector v.
ha, b (v)=INT((a·v+b)/r) equation (1)
where the element of vector a has a value extracted randomly from p-stable distributions typified by the Gaussian distribution, the scalar b has a value extracted randomly from the range of [0,r] in accordance with a uniform distribution and the function “INT( )” signifies round-off of decimal.
To add, the hash function used in the LSH is called a hash function sensitive to locality (hereinafter, termed a locally sensitive hash function) and is defined as will be described below. Where an object original space is S and a space of generated hash values is U, a locally sensitive hash set is defined when a hash function set H={h:S->U} satisfies conditions as below.
By using equation (1), a hash h[D[i]] of a dummy feature D[i] is prepared. The hash table is prepared as a table having an element (h[D[i]], D[i]). The similarity search using the LSH will be described later.
The thus prepared hash table is stored in the auxiliary data DB 114.
Then, the auxiliary data DB preparation terminal transmits to the client the data in the auxiliary data DB 114. Receiving the data, the client stores the data in the auxiliary data DB 103.
Next, a flow of registration process in the present embodiment will be described with reference to
The sensor 107 captures biometric information of a user (S401).
The feature extraction unit 101 extracts a feature x from the biometric information (S402). If a finger vein image is concerned, the method described in reference 5 can be used as feature extraction method.
The similarity search unit 102 reads the hash table {h[D[i]], D[i]} from the auxiliary data DB 103 (S403).
The similarity search on 102 searches the read-in hash table {h[D[i]], D[i]} and specifies a dummy feature resembling a feature x (S404). Here, as an example, the method called LSH described in references 6 and 8 is supposed to be used. Another method, for example, the AESA described in reference 7 may be used.
Firstly, by using equation (1), a hash value h[x] of the feature x is calculated. Subsequently, with respect to a set of hashes {h[D[i]]} in the read-in hash table {h[D[i]],D[i]}, a dummy feature coincident with the hash h[x] is searched. Generally, the dummy feature D[i] having the coincident hash value is one or more in number. These dummy features are called hash matched dummy features D*[i]. In the presence of a plurality of these hash matched dummy features, each hash matched dummy feature D*[i] is collated with a feature x. For matching, a method using cross-correlation, for example, may be used. A hash matched dummy feature having the maximal cross-correlation is determined as a dummy feature which is the most similar to the feature x. This is called the nearest neighboring dummy feature D*[i*]. The result of the similarity search is the index i* for the nearest neighboring dummy feature D*[i*].
The feature transform unit 104 reads a transform parameter R[i*] corresponding to the nearest neighboring dummy feature D*[i] (S405).
The feature transform unit 104 transforms the feature x by using the transform parameter R[i*] (S406). In finger vein biometrics authentication, for example, the similarity between two features may sometimes be calculated through cross-correlation. For matching algorithm in which the similarity is calculated on the basis the cross-correlation, an algorithm (correlation invariant random filtering) is known in which the similarity is calculated by applying a special transformation to the two features to keep them concealed and keep them unreturned (for details, see references 2 and 3). In this case, transform parameters R[i*] are given by random filter having individual pixel values in the form of random numbers. In the transformation process, the feature x first undergoes change of bases (such as number-theoretic transform or Fourier transform) and (data after the change of bases) is designated by X, and thereafter the X is multiplied by the random filter pixel-wise. The above is a typical example of transformation process but another method may be employed.
The feature T after transformation is used as a template T and an index i* for the nearest neighboring dummy feature D*[i*] is used for a group ID. Then, the template T and the group IDi* are transmitted to the server 220.
The server 120 registers the template T at a group having i* as ID in the template DB 122 (S407).
Next, 1:N authentication process in the present embodiment will be described with reference to
The feature extraction unit 101 extracts a feature y from the biometric information (502). As a method for feature extraction, the method described in reference 5 can be used in the case of, for example, a finger vein image.
The similarity search unit 102 reads the hash table {h[D[i]], D[i]} from the auxiliary data DB 103 (S503).
The similarity search unit 102 searches the read hash table {h[D[i]], D[i]} and specifies a dummy feature resembling a feature y (S504). Here, as an example, the method called LSH described in references 3 and 6 is supposed to be used but another method may be adoptable. For example, the AESA described in reference 7 may be used.
Firstly, by using equation (1), a hash value h[y] of the feature y is calculated. Subsequently, with respect to a set of hashes {h[D[i]]} in the read-in hash table {h[D[i]],D[i]}, a dummy feature coincident with the hash h[y] is searched. Generally, one or more dummy features D [i] having the coincident hash value may be present. These dummy features are called hash matched dummy features D*[i]. In the presence of a plurality of these hash matched dummy features, each hash matched dummy feature D*[i] is collated with the feature y. For matching, a method using cross-correlation, for example, may be used. A hash matched dummy feature having the maximal cross-correlation is determined as a dummy feature which is the most similar to the feature y. This is called the nearest neighboring dummy feature D*[i*]. The result of the similarity search is the index i* for the nearest neighboring dummy quantity D*[i*].
The feature transform unit 104 reads transform parameter R[i*] corresponding to the nearest neighboring feature D*[i*] (S505).
The feature transform unit 104 transforms the feature y by using the transform parameter R[i*] (S506). In finger vein biometrics authentication, for example, the similarity between two features may sometimes be calculated through cross-correlation. For matching algorithm in which the similarity is calculated on the basis of the cross-correlation, an algorithm (correlation invariant random filtering) is known in which the similarity is calculated by applying a special transformation to the two features to keep them concealed and keep them unreturned (for details, refer to references 2 and 3). In this case, transform parameters R[i*] are given by a random filter having individual pixel values in the form of random numbers. In the transformation process, the feature y first undergoes change of bases (such as number-theoretic transform Fourier transform) and (data after the change of bases) is designated by Y, and thereafter the Y is divided by the random filter pixel-wise. The above is a typical example of transformation process but another method may be employed.
An index i* for the nearest neighboring dummy feature D*[i*] is used as a group ID. Then, the transformed feature Y and the group IDi* are transmitted to the server 120.
The server 120 performs a 1:N matching between the transformed feature V and templates contained at a group having i* as ID in the template DB 122 (S507). In finger vein biometrics authentication, for example, the similarity between two features may sometimes be calculated through cross-correlation. For matching algorithm in which the similarity is calculated on the basis of the cross-correlation, an algorithm (correlation invariant random filtering) is known in which the similarity is calculated by applying a special transformation to the two features to keep them concealed and keep them unreturned (for details, see references 2 and 3). In this case, transform parameters are common in the group i* and therefore, the 1:N matching can be realized in such way that 1:1 matching between each of the templates inside the group i* and the transformed feature V is executed once and an ID for a template having the highest correlation is outputted as the result.
According to the present embodiment, during authentication, allotment to a particular group executed by performing similarity search to the dummy feature on the side of client 100, so that the number of templates subject to 1:N matching on the side of the server 120 is allowed to narrow down drastically, thereby ensuring that the load processed on the server side can be educed and the time to process the 1:N matching can be decreased to great extent.
As has been set forth so far, according to the present embodiment, in the 1:N authentication, the load on the server side can be reduced and speedup of the process can be achieved. Besides, since not only features of biometric information belonging to a specified individual but also dummy features are held in the client 100, problems of privacy and security do not matter.
The foregoing embodiment can be applicable to an arbitrary application in which user authentication the basis of biometric information.
For example, widespread application can be attained to, for example, information access control in an internal office network, personal confirmation in internet banking system or ATM, log-in to a Web site toward members, personal authentication during entrance to a protective area, log-in in personal computers, and so on.
The specification and drawings are, accordingly, to be regarded in an illustrative rather that a restrictive sense. It will, however, be evident that various modifications and changes may be made thereto without departing from the spirit and scope of the invention as set forth in the claims.
Number | Date | Country | Kind |
---|---|---|---|
2009-138778 | Jun 2009 | JP | national |
Number | Name | Date | Kind |
---|---|---|---|
5841888 | Setlak et al. | Nov 1998 | A |
5845005 | Setlak et al. | Dec 1998 | A |
5892838 | Brady | Apr 1999 | A |
6118890 | Senior | Sep 2000 | A |
6181807 | Setlak et al. | Jan 2001 | B1 |
6661908 | Suchard et al. | Dec 2003 | B1 |
6941003 | Ziesig | Sep 2005 | B2 |
6954553 | Ikegami | Oct 2005 | B2 |
6985583 | Brainard et al. | Jan 2006 | B1 |
7046829 | Udupa et al. | May 2006 | B2 |
7278028 | Hingoranee | Oct 2007 | B1 |
7330572 | Uchida | Feb 2008 | B2 |
7502467 | Brainard et al. | Mar 2009 | B2 |
7512255 | Kakadiaris et al. | Mar 2009 | B2 |
7580551 | Srihari et al. | Aug 2009 | B1 |
7773784 | Boult | Aug 2010 | B2 |
8031981 | Peirce | Oct 2011 | B2 |
20020118864 | Kondo et al. | Aug 2002 | A1 |
20020152034 | Kondo et al. | Oct 2002 | A1 |
20030039380 | Sukegawa et al. | Feb 2003 | A1 |
20030103652 | Lee et al. | Jun 2003 | A1 |
20040062423 | Doi | Apr 2004 | A1 |
20040230810 | Hillhouse | Nov 2004 | A1 |
20040255168 | Murashita et al. | Dec 2004 | A1 |
20050008200 | Azuma et al. | Jan 2005 | A1 |
20050152583 | Kondo et al. | Jul 2005 | A1 |
20050238214 | Matsuda et al. | Oct 2005 | A1 |
20050249385 | Kondo et al. | Nov 2005 | A1 |
20060021003 | Fisher et al. | Jan 2006 | A1 |
20060158307 | Lee et al. | Jul 2006 | A1 |
20060204058 | Kim et al. | Sep 2006 | A1 |
20060210126 | Cho | Sep 2006 | A1 |
20070003112 | Awatsu et al. | Jan 2007 | A1 |
20070031010 | Sukegawa et al. | Feb 2007 | A1 |
20070036400 | Watanabe et al. | Feb 2007 | A1 |
20070061590 | Boye et al. | Mar 2007 | A1 |
20070118758 | Takahashi et al. | May 2007 | A1 |
20070253624 | Becker | Nov 2007 | A1 |
20070276853 | Hamza | Nov 2007 | A1 |
20070286465 | Takahashi et al. | Dec 2007 | A1 |
20080013803 | Lo et al. | Jan 2008 | A1 |
20080037833 | Takahashi et al. | Feb 2008 | A1 |
20080065900 | Lee et al. | Mar 2008 | A1 |
20080072063 | Takahashi et al. | Mar 2008 | A1 |
20080101658 | Ahern et al. | May 2008 | A1 |
20080178002 | Hirata et al. | Jul 2008 | A1 |
20080178008 | Takahashi et al. | Jul 2008 | A1 |
20080222426 | Schrijen et al. | Sep 2008 | A1 |
20080273770 | Kohout | Nov 2008 | A1 |
20080298642 | Meenen | Dec 2008 | A1 |
20090006855 | Tuyls et al. | Jan 2009 | A1 |
20090007257 | Hirata et al. | Jan 2009 | A1 |
20090034805 | Perlmutter et al. | Feb 2009 | A1 |
20090070860 | Hirata et al. | Mar 2009 | A1 |
20090226053 | Matsuda et al. | Sep 2009 | A1 |
20090310830 | Bolle et al. | Dec 2009 | A1 |
20100040265 | Iizuka | Feb 2010 | A1 |
Number | Date | Country |
---|---|---|
2007-293807 | Nov 2007 | JP |
Number | Date | Country | |
---|---|---|---|
20100315201 A1 | Dec 2010 | US |