Mobile authentication system and method based on Application on Voiceprint Recognition, head portrait identification and location-based service
The present invention relates to network communication field, relate in particular to the method and system that a kind of use mobile terminal device (as mobile phone, panel computer etc.) carries out authentication, further relate to a kind of method and system of carrying out authentication of combining Application on Voiceprint Recognition, head portrait identification, location technology multi-modal information.
Authentication system has very widely application in a lot of fields, and for example: enterprise's work attendance, register etc. in classroom.Not only efficiency is not high but also the reliability of authentication also can't guarantee for traditional punched-card machine, and fingerprint attendance machine has recently adopted more reliable security mechanism but still can't solve the problem of efficiency.In the face of efficiency, traditional method can only solve by buying more equipment, and traditional equipment can only be fixed on certain special position and can only provide service for a specific class event, adopts such solution obviously can significantly increase cost.
We find that utilizing existing mobile terminal is a kind of solution of very simple economy, utilizes these equipment directly to make personnel costs drop to minimum.Existing many location-based services of registering in current mobile device (by the Ru“ street " etc.), but these application do not comprise the function of authentication, just mean that these application can only be confined to amusement yet, can't carry out serious business application.
Traditional Time Attendance Device has many shortcomings, such as: the deployed with devices cost is high, verification mode is single and the portable shortcomings such as poor.Just because of these problems, caused the limitation of legacy equipment application scenarios.
Summary of the invention
This present situation for a change, we merge the system of registering of the multi-modal high precision identity federation verification system such as voice, image and position-based information, design the mobile authentication system of a new generation.This system has efficiently, accurate, the many-sided advantage of economic dispatch, so being not limited only to school of enterprise, its application scenarios registers, for example: member user's authentication in service industry, the fixed point of taking an oath in orienteering match is registered, the convict be released on bail for medical treatment or supervision of residence during the regularly authentication etc. of fixed point.
The mobile authentication system that the system of registering of the multi-modal high precision identity federation verification system such as a kind of voice, image and position-based information of the invention provides merges comprises:
Voice vocal print training unit, for extracting the acoustic feature of voice signal, training generates user's voice sound-groove model, and this sound-groove model will be for user's speech recognition checking;
Voice vocal print authentication unit, for the similarity of user's relatively input voice and user's voice sound-groove model, the identity of the contrast verification user by this unit;
Face's detection and identify unit, by detecting the feature of user face face, the judgement face location, return to facial information;
Face's similarity comparing unit, by contrasting the feature of two face images, the similarity of judgement face, and the comprehensive similarity of output integrated judgement;
The comprehensive verification unit, for the comprehensive assessment to face's similarity comparing unit, voice vocal print authentication unit and geographical location information, export final assessment result.
The present invention also provides the multi-modal high precision identity federation verification method of a kind of voice, image, position-based information etc., it is characterized in that on the basis of the personal identification method that adopts traditional voice, identity, face, realized multi-modal high precision combined identity certification, the method has been avoided single capacity authentication inaccuracy, and can carry out efficiently, rapidly authentication.This combined identity certification method specifically comprises the steps:
The first step: on the first logging status of active user, by sample user's face image of terminal, by the eigenwert of this face of detection and identify unit inspection of face, and deposit among this user's database.By sample user's sound of terminal, term speech line training module train this user's of generation voice vocal print engine to the sound of input again;
Second step: on the first logging status of active user, and, in the situation in the unregistered place of registering, to system, submit to future by the address of registering, system is put into system database by this user profile;
The 3rd step: after the active user has produced voice vocal print engine, the user submits own current speech sample to system, system creation initialization speech recognition engine, and speech samples input speech recognition engine, engine will be exported the voice similarity, and be recorded in the result database of this checking;
The 4th step: after the active user has registered face, the user submits own current face's sampling to system, system creation initialization face detection engine, and, face's sample input face detection engine, engine is exported face image.Then, system creation generates the face recognition engine, and the face that the face registered in database and active user upload is input among the face recognition engine, and engine will be exported face's similarity degree, and be recorded in the result database of this checking;
The 5th step: in the process of being verified the user, terminal will gather current geographic position information, and after successfully gathering, background system will be recorded in geographical location information in the result database of this checking;
The 6th step: after the user has submitted all authorization informations to, system will be carried out joint qualification to information such as voice, image, positions, adopt the weighting appraisal procedure, accuracy of identification to each is weighted, and the weighted value of all is carried out to product calculation, if reach the approval threshold value, and the individual event resolution must not set lowest threshold lower than it, can think that joint qualification passes through.
In above-mentioned mobile authentication system, be not limited to weights for the mode that realizes joint qualification and add up, any other mode that can be combined judgement to multinomial recognition result may be used to realize joint qualification.
The accompanying drawing explanation
Fig. 1 is the multi-modal high precision identity federation verification system general structure of voice, image, position-based information etc.
In order to make technical matters to be solved by this invention, technical scheme and beneficial effect clearer, below in conjunction with drawings and Examples, the present invention will be described in detail.It should be noted that specific embodiment described herein, only in order to explain the present invention, is not intended to limit the present invention.
Fig. 1 has provided the multi-modal high precision identity federation verification system general structure of voice, image, position-based information etc., and this system mainly is comprised of face's detection and identify unit, voice vocal print training unit, voice vocal print authentication unit, face's similarity comparing unit, geography information sampling unit, comprehensive verification unit.
Face's detection and identify unit: by detecting the feature of user face face, the judgement face location, extract facial information.
Face's similarity comparing unit: by contrasting the feature of two face images, the similarity of judgement face, and the comprehensive similarity of output integrated judgement;
Voice vocal print training unit: for extracting the acoustic feature of voice signal, training generates user's voice sound-groove model, and this sound-groove model will be for user's speech recognition checking;
Voice vocal print authentication unit: for the difference of user's relatively input vocal print and user's voice sound-groove model, the identity of the contrast verification user by this unit;
Geography information sampling unit: for the geography information position of the current terminal of sampling, by the spherical co-ordinate computing, obtain the distance with the place of registering, carry out the checking in geographic position;
Comprehensive verification unit: for the comprehensive assessment to face's similarity comparing unit, voice vocal print authentication unit and geographical location information, export final assessment result;
Described joint qualification algorithm includes but not limited to following several mode:
(1) adopt the weighting appraisal procedure: the accuracy of identification to each is weighted, and the weighted value of all is carried out to product calculation, if reach the approval threshold value, and the individual event resolution must not set lowest threshold lower than it, can think that joint qualification passes through.
ibe the identification weights of i item, d
ibe the resolution of i item, v
ibe the weighting resolution of i item, δ
ibe the minimum requirements resolution of i item, θ is the associating resolution.
(2) 0,1 criterion: can only export 1 and at 0 o'clock for each authentication unit, we adopt 0,1 to judge, must all resolutions of thinking be 1 o'clock, and final the result could be 1.
ibe the weighting resolution of i item, θ is the associating resolution.
(3) three threshold determination methods: for authentication unit can export 1 and 0 and can set the checking threshold value the time, we adopt three subthresholds to judge and obtain more accurate resolution.Verified by the initial setting threshold value for the first time, if be verified, improve threshold value and again verify; If checking is not verified again by reducing threshold value.Triplicate checking, the result of each checking is that 1 or 0, three result forms the binary number of 3 as the resolution of this project checking.Because initial threshold value has the greatest impact to result, so, using the result of front as a binary number high position, last the result is low level, the resolution obtained is d
i.After having obtained like this individual event resolution, can obtain the associating resolution:
ibe the resolution of i item, θ is the associating resolution.
Below further take mobile register to register as example describe.The registration of being registered of traditional mode of checking card or signing of adopting of registering, if it is numerous to register, not only can bring crowdedly, for the tissue side of registering, also needs to bear a lot of human costs and equipment cost.Adopt native system can utilize the person's of registering mobile phone terminal to carry out registering of accurate quick, thereby improved the efficiency of registering, reduced the cost of registering.
In the time of first use native system, the user need to be registered personal information, geography information and be extracted the face's sample of oneself and the sound-groove model of training oneself.When being registered, the user need to submit current face image and speech samples to system, and system can be identified according to the facial model retained in database and sound model, then, according to the geography information of user terminal, carries out the joint qualification user identity.The assurance of recognition of face and speech recognition higher degree the people that registers be really identity, the checking of geography information has guaranteed the accuracy in geographic position, thereby has reached the combined identity certification of pin-point accuracy.
The mobile authentication system that the system of registering of the multi-modal high precision identity federation verification system such as voice of the present invention, image and position-based information merges has advantages of as follows: the equipment of whole system, comprise Authentication server, mobile authentication terminal and system, only need can use at server and mobile device deploy software, save equipment cost.Used Application on Voiceprint Recognition, recognition of face and positional information associating dynamic authentication, the result is reliable more comprehensively.Use handheld device to support the large-scale groups simultaneous verification, significantly improved verifying speed.Whole proof procedure is all completed automatically by server, easy to operate.
The foregoing is only preferred embodiment of the present invention, be not limited to the present invention, all any modifications of doing within principle of the present invention and spirit, be equal to and replace and improvement etc., within all just being included in protection scope of the present invention.And for example: the joint qualification of combining geographic information and face recognition, the joint qualification of combining geographic information and speech recognition and other joint qualification combinations.