CN201955775U - USB identity authentication system based on voiceprint identification - Google Patents

Abstract

The utility model relates to a USB (Universal Serial Bus) identity authentication system based on voiceprint identification, belonging to the technical field of computer information safety. The USB identity authentication system comprises a user terminal and an embedded voiceprint identification hardware system based on the USB, wherein the user terminal is connected with the embedded voiceprint identification hardware system based on the USB by a USB bus, the embedded voiceprint identification hardware system based on the USB comprises a central processing unit used for completing voiceprint identification operations of the authentication, and the central processing unit is respectively connected with a voice acquisition unit, a data storage unit, a power supply management module, a status indication module and a USB interface used for connecting the user terminal. The USB identity authentication system in the utility model has the advantage of high security, and can be expanded to the field of network bank, electronic commerce, e-government, etc. by providing third party application and service voiceprint authentication interfaces.

Description

USB identity authorization system based on Application on Voiceprint Recognition

Technical field

The utility model relates to the computer information safety technique field, especially relates to USB identity identifying method and system thereof based on Application on Voiceprint Recognition.Mode by Application on Voiceprint Recognition authenticates user's identity, improves the security of computing machine.

Background technology

By consulting related data and relevant patent, mainly quote following patent as proof:

The open date of name of the country publication number

1. Chinese CN 101340285A 2009.1.7

2. Chinese CN 101819614A 2010.9.01

Patent 1 (CN 101340285A) application discloses the method and system of utilizing fingerprint USBkey to carry out authentication.Finger print information typing fingerprint USBkey with the user preserves in advance, generate digital certificate, each digital certificate is stored on the fingerprint USBkey with corresponding encrypted private key, when the user communicates by letter with processing platform, the fingerprint USBkey of user side is the authentication fingerprint earlier, after the data that needs sent with corresponding private key are carried out digital signature again, be sent to processing platform and carry out authentication.Patent 2 (CN101819614A) discloses a kind of system and method that utilizes voice verification USBKey to strengthen network transaction safety.CA and system for handling network transactions are confirmed user identity by the digital certificate in the voice verification USBKey, the transaction data that the user imports when carrying out online transaction is sent to voice verification USBKey by user terminal and is carried out data signature, voice verification USBkey carries out voice broadcast to operational order and the operating result that receives, improve the operating experience that the user uses voice verification USBKey, assisting users is judged transaction data, and to carry out the transaction data of data signature consistent with sending to voice verification USBKey, improves the security of online transaction.Voice verification USBkey has also increased fingerprint identification module simultaneously, utilizes user's finger print information to carry out authentication.

Present authentication means based on personal computer and network, comprise PKI (Public Key Infrastructure, Public Key Infrastructure)/CA (Certificate Authority, authentication center), smart card and USB certificate token etc., finally do not break away from dependence, all the ubiquity critical defect " recognizing thing, do not recognize people, can not discern subjective consciousness " to " thing ".Biological identification technology utilizes people's biological characteristic authentication user identity, and user's physical identity is corresponding with digital identity, becomes reliable identity authentication means.Although above-mentioned patent combines fingerprint identification technology and USBkey, improve the security of network trading, but mainly have following problem for the technical requirement aspect that solves the application: above-mentioned patent is mainly used in the safety issue that improves network trading, need link to each other with network trading platform with the CA center.

Summary of the invention

The utility model provides a kind of USB identity identifying method and system thereof based on Application on Voiceprint Recognition.Purpose is to use user's voiceprint that the user is carried out authentication, avoids forgeing or revealing because of password the safety problem of initiation.The technical scheme that the utility model is taked is:

Comprise user terminal and based on the embedded Application on Voiceprint Recognition hardware system of USB, user terminal links to each other by usb bus with embedded Application on Voiceprint Recognition hardware system based on USB, the central processing unit that should comprise the Application on Voiceprint Recognition computing that is used to finish authentication, this central processing unit based on embedded Application on Voiceprint Recognition hardware system of USB respectively with voice collecting unit, data storage cell, power management module, state indicating module and the USB interface that is used to be connected user terminal connect.

A kind of embodiment of the utility model is, comprises program and vocal print database storing space module, secure memory space module, open storage space module in the data storage cell based on the embedded Application on Voiceprint Recognition hardware system of USB; Wherein program code stored, the user's of program and vocal print database storing space module voiceprint template and voice answer, secure data storage module and open storage space module adopt the high-capacity FLASH storer to realize, wherein openly the storage space module user can carry out free operant to data wherein, and the operation of secure data storage module then needs the vocal print of authenticated.

This patent is the portable product at the user personal computer information security management, only needs user terminal and based on the embedded Application on Voiceprint Recognition hardware system of USB.Compare with fingerprint identification technology, the Application on Voiceprint Recognition technology is easy to use, is easier to accept.The voice that contain the vocal print feature obtain convenient and nature, and voiceprint extraction can be finished unconsciously, user's acceptance level height.Obtaining of vocal print do not need very complicated hardware, and a microphone gets final product, and more need not extra sound pick-up outfit when using communication facilities.The Application on Voiceprint Recognition technology is fit to remote identity and confirms.Only need microphone, phone or a mobile phone just can pass through network (communication network or internet) and realize Telnet.In addition, the complexity of Application on Voiceprint Recognition algorithm is low, can cooperate other measures, and as carrying out content discriminating etc. by speech recognition, the two natural combination can realize dual safety certification, further strengthens security intensity.

Advantage of the present utility model is, uses user's voiceprint that the user is carried out authentication, avoids forgeing or revealing because of password the safety problem of initiation.Can related every authentication based on the USB identity authorization system of Application on Voiceprint Recognition, Application on Voiceprint Recognition algorithm in conjunction with speech recognition, each pronunciation is controlled by prompting problem at random, prevent recording, imitation, voiceprint authentication method realizes that in the embedded Application on Voiceprint Recognition hardware system based on USB identification is accurate, the vocal print versus speed is fast, can satisfy the requirement of system real time fully, and user's voiceprint is stored in the embedded Application on Voiceprint Recognition hardware system based on USB, and is safer, reliable.Adopt flash media, have powerful shatter-resistant antidetonation character.Can use as the vocal print encrypted U disk, even lose based on the embedded Application on Voiceprint Recognition hardware system of USB, being stored in its secure data spatial data can not be opened, and has higher security.The utility model can also provide third party's application and service voiceprint interface, can expand to fields such as the Internet bank, ecommerce, E-Government.

Description of drawings

Fig. 1 is a principle schematic of the present utility model.

Fig. 2 is the interface block diagram of central processing unit DSP of the present utility model and AIC23.

Fig. 3 is the power management module circuit diagram of user terminal system when being PC.

Fig. 4 is the process flow diagram of the utility model based on the USB identity identifying method of Application on Voiceprint Recognition.

Fig. 5 is the process flow diagram of user's of the present utility model voiceprint typing.

Fig. 6 is the process flow diagram of voiceprint process of the present utility model.

Fig. 7 is the process flow diagram that the utility model cancellation user terminal data vocal print is encrypted.

Fig. 8 is the process of TMS320VC5509 data transmission during as USB device.

Fig. 9 is the AIC23 circuit theory diagrams.

Embodiment

The utility model discloses a kind of USB identity identifying method and system thereof based on Application on Voiceprint Recognition.Comprise user terminal 101 and based on the embedded Application on Voiceprint Recognition hardware system 103 of USB, as shown in Figure 1 based on the USB identity authorization system of Application on Voiceprint Recognition.User terminal 101 links to each other by usb bus 102 with embedded Application on Voiceprint Recognition hardware system 103 based on USB.User terminal 101 can be that personal computer, mobile phone, palm PC PDA (Personal Digital Assistant) etc. have processing power and other equipment by the USB main interface are set; Embedded Application on Voiceprint Recognition hardware system 103 based on USB comprises central processing unit 201, speech signal collection module 202, system data memory module 203, power management module 204, state indicating module 205, USB interface 206.USB interface 206 is used to connect user terminal.Said system is a core with central processing unit 201, control speech signal collection module 202, system data memory module 203, power management module 204, state indicating module 205, USB interface 206.Wherein, central processing unit 201 adopts the TMS320VC5509DSP digital signal processor that has usb interface module of TI company, the Application on Voiceprint Recognition algorithm that is used to authenticate, the interface operation of control peripheral chip.

Speech signal collection module 202 is made up of Mike and programmable-gain, sampling rate adjusting A/D converter, is finished the gain control and the digitized processing of voice signal by central processing unit 201 controls.Fig. 2 has provided the interface block diagram of TMS320VC5509DSP and AIC23.

System data memory module 203 comprises program and vocal print database storing space module, secure memory space module, discloses storage space module.Wherein program code stored, the user's of program and vocal print database storing space module voiceprint template and voice answer.Secure data storage module and open storage space module adopt the high-capacity FLASH storer to realize, wherein openly the storage space module user can carry out free operant to data wherein, and the operation of secure data storage module then needs the vocal print of authenticated.

The power management module 204 of system is realized by the power conversion chip circuit usually, so that the operating voltage of devices such as central processing unit, FLASH storer to be provided.System's dispense with outer connecting power can obtain voltage from the usb bus of user terminal by USB interface.For different user terminals, the voltage that its usb bus provided is also different, can use the proframmable linear voltage regulator, as the MAX667 of U.S. MAXIM company production, usb bus voltage is carried out voltage transitions, make voltage satisfy the voltage request of power conversion chip.Fig. 3 has provided the power management module circuit diagram of system when user terminal is PC.When user terminal was PC, the output voltage on the usb bus was 5V at this moment, the conversion that realizes voltage specially for the power conversion chip TPS767D301 of digital signal processor power supply that can directly adopt TI company to produce.Its standard of this chip is input as 5V, and is identical with the usb bus output voltage, standard is output as+3.3V and+1.6V.Its+3.3V output not only is the central processing unit power supply, also for own in the system+the 3.3V device provides power supply.1.6V for TMS320VC5509DSP provides operating voltage.

System state indicating module 205 is made up of three pilot lamp and a hummer, the state that indication mechanism is current and in the process of user's voiceprint typing, and the prompting user operates accordingly.

USB control module that dsp chip TMS320VC5509 is integrated, can finish and the usb host system between read-write operation, have and need not to add logical circuit, advantage such as easy to use.Here the USB module realizes on the sheet of USB interface 206 use TMS320VC5509, can finish the data communication between DSP and the user terminal.The process of data transmission when Fig. 8 has provided TMS320VC5509 as USB device.In the IN affairs, SIE (serial interface engine) receives data from UBM (buffer manager for use), gives user terminal (main frame) after converting serial data stream to.In the OUT affairs, SIE converts the serial data of main frame to the parallel form of UBM.UBM transmits data between SIE and impact damper RAM.UBM with data transmission before the SIE, the dma controller of CPU or USB must be with deposit data in impact damper RAM.When CPU or dma controller have been ready to that data are moved into the storer of DSP, must wait for that UBM moves on to impact damper RAM to data from SIE.

The utility model based on the process flow diagram of the USB identity identifying method of Application on Voiceprint Recognition as shown in Figure 4.For the user of first use based on the USB identity authorization system of Application on Voiceprint Recognition, at first the embedded Application on Voiceprint Recognition hardware system based on USB to be inserted user terminal, user terminal is provided with control program, according to the prompting typing user's of user terminal voiceprint, and it is stored.On user terminal, the selected vocal print enciphered data that needs the vocal print authentication realizes that the vocal print encryption is provided with then.During visit data, client terminal system can judge at first whether these data are encrypted by vocal print, if do not encrypted by vocal print, then the user can directly visit to the user, and carries out corresponding operating, as copy, deletion etc. on user terminal.If data are encrypted by vocal print, the embedded Application on Voiceprint Recognition hardware system of the USB that client terminal system uses in the time of can pointing out the user to insert the vocal print typing, and the input authentication voice carry out voiceprint.If user's voiceprint authentication success, then the user can conduct interviews and operates the vocal print ciphered data, otherwise, can't visit and operate.

Vocal print encryption and two processes of voiceprint of mainly comprising data based on the USB identity identifying method of Application on Voiceprint Recognition.In the vocal print ciphering process of data, the process flow diagram of user's voiceprint typing as shown in Figure 5.The user when entry information,

The problem that needs voice answering when (a) user terminal is pointed out the typing of user's vocal print, 4 or 5 problems are arranged usually, these problems normally with some subscriber-related essential informations, as birthday of own or father and mother, phone etc., and the form of voice answer is generally continuous numeric string form.

(b) user imports the voice answer on based on the embedded Application on Voiceprint Recognition hardware system of USB.

(c) gather user's voice based on the embedded Application on Voiceprint Recognition hardware system of USB, and these voice answers are carried out characteristic parameter extraction, form voice answer template and speaker's voiceprint template, with the template stores that forms in voice print database storage space module based on the embedded Application on Voiceprint Recognition hardware system of USB.

(d) by USB interface the signal of user's voiceprint typing success is transferred to user terminal based on the embedded Application on Voiceprint Recognition hardware system of USB, by user terminal prompting voiceprint typing success.

The user is when the typing voiceprint, and its step (c) needs to form user's voice answer template and speaker's voiceprint template, and its specific implementation algorithm is as follows:

(i) formation of voice answer template

Step 1: the embedded Application on Voiceprint Recognition hardware system based on USB is gathered the user's voice answer, and pre-service is carried out in the voice answer of gathering, and pre-service comprises the cutting apart of branch frame, windowing, consecutive numbers word string, the pre-emphasis of signal.

Step 2: user speech answer numeric string is split into D digital O={O ¹, O ²... O ^D, each numeral is extracted respectively based on the Gammachip characteristic parameter of examining PCA.Leaching process is as follows:

If Xt (p), p=0..., N-1, t=1 ... T is the pretreated signal of process, and N is counting of every frame, and T is the frame number that signal divides frame.

At first to every frame voice signal X _t(p) the FFT conversion of dispersing, delivery square obtain discrete power spectrum S _t(n), n=0 ..., N/2.

Y _t(n)＝FFT(X _t(p))，S _t(n)＝|Y _t(n)| ² (1)

Calculate S then _t(n) by each H _m(n) performance number that obtains after is promptly calculated S _t(n) and H _m(n) sum of products on each discrete point in frequency obtains M parameter P of every frame signal _t(m), m=0 ..., M-1, t=1 ..., T.

$P_t\left(m\right)=\underset{n=0}{\overset{N/2}{\mathrm{\Σ}}}{H}_m\left(n\right)S_t\left(n\right)---\left(2\right)$

H wherein _m(n) be several Gammachip bank of filters H that in the spectral range of voice, is provided with _m(n), m=0 ..., M-1; N=0 ..., N/2, M are number of filter, N is counting of a frame voice signal.H _m(n) can be by the impulse response g of Gammachip bank of filters _c(t) power spectrum obtains.

g _c(t)＝at ^n-1exp(-2πbERB(f _r)t)·exp(j2π·f _rt+jclnt+φ)

Time t＞0 in the following formula, a is an amplitude, n and b be the parameter of gamma function distribution situation for a change, ERB (f _r) be that centre frequency is f _rThe value ERB (f of ERB of auditory filter _r)=24.7+0.108f _r

Obtaining the Gammachip characteristic parameter P of all frame signals _t(m), m=0 ..., M-1, t=1 ..., after the T, adopt the multivariate statistical analysis of nuclear PCA, to P _mCarry out dimension-reduction treatment.Process is as follows:

1. to P _t(m), m=0 ..., M-1, t=1 ..., the centralization of T sequence vector.The computation of mean values vector

Obtain the sequence vector P ' after the centralization _t(m)=P _t(m)-and μ, t=1 ..., T.

2. calculate P _t(m) nuclear matrix

Wherein J is the unit vector of M*1, and K is two parameter P _t(m) and P _i(m) nuclear matrix between is

I=1 ... T, m=1 ... T.Wherein σ is the nuclear width of gaussian kernel function.

3. preceding k the eigenvalue of maximum λ of compute matrix K _i, i=1...k and characteristic of correspondence vector v _i, i=1...k.

4. the characteristic parameter behind the computational transformation

Wherein ${\mathrm{\α}}^j=\frac{1}{\sqrt{{\mathrm{\λ}}_j}}{v}_j,j=1,...k$

After nuclear PCA conversion, the dimension of new feature vector is the k dimension.

Step 3: with each digital pairing characteristic parameter respectively with embedded Application on Voiceprint Recognition hardware system based on USB in the speech model of the good 0-9 numeral of the precondition stored carry out the score coupling, find the pairing numeral of model of each digital score maximum, as the template stores of voice answer in embedded Application on Voiceprint Recognition hardware system based on USB.

If d digital O ^dPairing characteristic parameter is Calculate respectively

Speech model Λ with the good 0-9 numeral of the precondition of storing in the embedded Application on Voiceprint Recognition hardware system based on USB _i, i=0 ..., 9 coupling, wherein Λ _i, i=0 ..., 9 adopt the model training method based on continuous Hidden Markov.If d digital O ^dPairing characteristic parameter

With the score maximum of q model, promptly

Then this numeral is identified as digital q, makes s _b(j)=q.At this moment, user terminal prompting user needs b the pairing voice answer of problem template of voice answering to be { s _b(j) }, b=1 ..., Q, j=1 ... W _b, need the Issue Totals of voice answering, W when wherein Q is the typing of user terminal prompting user vocal print _bIt is the number that the answer of b problem user speech comprises numeral.

(2) formation of user's vocal print template

Step 1: the user speech answer of system to gathering based on the embedded Application on Voiceprint Recognition hardware system of USB, carry out pre-service, comprise branch frame, windowing, end-point detection, the pre-emphasis of signal.

Step 2:, extract Gammachip characteristic parameter based on nuclear PCA to pretreated data

(step 2 that leaching process generates with sound template).

Step 3: utilize Gammachip characteristic parameter based on nuclear PCA

Generate GMM-UBM model Λ _s={ w _l, μ _l, ∑ _l, l=1 ..., R is with the mean vector μ of each degree of mixing _lBe joined together to form GMM-UBM super vector X=[μ ₁μ ₂... μ _R] ^T

Step 4: with X=[μ ₁μ ₂... μ _R] ^TAs input, training is based on the support vector machine speaker model of GMM-UBM super vector.The structural classification function Make that sample X is carried out the error probability of branch time-like is as far as possible little.

Here find the solution dual equation (1) and (2), and, obtain α according to Karush-Kuhn-Tucker (KKT) condition _i ^*And parameter b ^*

$\max H\left(a\right)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{a}_i-\frac{1}{2}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{y}_i{y}_j{a}_i{a}_{j}K(x_i,x_j)---\left(1\right)$

$\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{y}_i{a}_i=0,a_i\≥0,i=\mathrm{1,2},...n---\left(2\right)$

K (x wherein _i, x _j) be the kernel function of two eigenvectors among the sample X.y _iBe sample x _iPairing category label, a _i〉=0 is the Lagrange coefficient.

With α _i ^*≠ 0 pairing sample vector x _j, category label y _jAnd parameter b ^*Sound-groove model as the speaker is stored.

Based on the process flow diagram of voiceprint process in the USB identity identifying method of Application on Voiceprint Recognition as shown in Figure 6.The user when carrying out authentication,

The problem that user speech was answered when (a) user terminal was randomly drawed a vocal print typing, the prompting user imports the voiceprint voice;

(b) user imports the voice answer on based on the embedded Application on Voiceprint Recognition hardware system of USB;

(c) gather user's authentication voice based on the embedded Application on Voiceprint Recognition hardware system of USB, extract the characteristic parameter of authentification of user voice, then characteristic parameter is compared with voice answer template and speaker's voiceprint template respectively, calculate two matching values;

(d) if the user's voice answer is correct, and user's voiceprint is compared successfully, then user's voiceprint passes through, otherwise, the failure of user's voiceprint, if the number of retries of user's voiceprint does not reach stipulated number, the user randomly drawed one the problem with voice answering when then user terminal can be again from the vocal print typing, wait for that the user carries out voice answering, the vocal print to the user in the number of times of regulation authenticates again; Otherwise, if user's authentification failure still in the number of retries of regulation, then could be by locked based on the embedded Application on Voiceprint Recognition hardware system of USB.

The user is when carrying out voiceprint, and its step (c) comprises comparison of voice answer template and the comparison of speaker's voiceprint template.

(i) voice answer template comparison

The problem that user speech was answered when if user terminal was randomly drawed b vocal print typing, the authentification of user voice are after pre-service, feature extraction, and authentification of user speech digit string is split into D _rIndividual numeral, the Gammachip characteristic parameter based on nuclear PCA of establishing d numeral is

D _rFor digital number, T cut apart in the authentification of user voice _dIt is the number of d divided numerical characteristic parameter.If the number D of the numeral that the authentification of user voice comprise _rBe not equal to the number W that the answer of b problem user speech comprises numeral _b, think that then the user's voice answer is incorrect.Otherwise, according to voice answer template { s _b(d) }, d=1 ... W _bFind the model of b pairing all the voice answers of problem, calculate the characteristic parameter that each cuts apart numeral successively With model Λ (s _b(d)) matching score.If each cuts apart the matching score of numeral all less than preset threshold, think that then the user's voice answer is correct, otherwise the user's voice answer is incorrect.

(ii) speaker's voiceprint template comparison

If

Be the Gammachip characteristic parameter that the authentication voice obtain, the model parameter α that produces during according to the typing of user's voiceprint based on nuclear PCA after pre-service, feature extraction _i ^*, x _i, y _iAnd parameter b ^*, the computational discrimination function

Value.

Be two kernel functions between the vector.If

The authentication voice of then thinking the user are that the user is said, and user's voiceprint is compared successfully.Otherwise

User's voiceprint comparison failure.

Fig. 7 is the process flow diagram that cancellation user terminal data vocal print is encrypted.If the vocal print encryption of cancellation user terminal system data, the embedded Application on Voiceprint Recognition hardware system of the USB that the user uses in the time of need inserting the vocal print typing, and will carry out voiceprint.If voiceprint passes through, then selecting in the control program of user terminal needs cancellation vocal print ciphered data, restarts user terminal then, and at this moment, the vocal print of this partial data is encrypted and just cancelled, and the user can directly conduct interviews to data without voiceprint.Otherwise the user can retry, as long as voiceprint success in the number of retries of regulation, the user can cancel the vocal print of data and encrypt, otherwise, will be based on the embedded Application on Voiceprint Recognition hardware system of USB by locked.

Claims (2)

1. USB identity authorization system based on Application on Voiceprint Recognition, it is characterized in that, comprise user terminal and based on the embedded Application on Voiceprint Recognition hardware system of USB, user terminal links to each other by usb bus with embedded Application on Voiceprint Recognition hardware system based on USB, the central processing unit that should comprise the Application on Voiceprint Recognition computing that is used to finish authentication, this central processing unit based on embedded Application on Voiceprint Recognition hardware system of USB respectively with voice collecting unit, data storage cell, power management module, state indicating module and the USB interface that is used to be connected user terminal connect.

2. the USB identity authorization system based on Application on Voiceprint Recognition according to claim 1, it is characterized in that, comprise in the data storage cell based on the embedded Application on Voiceprint Recognition hardware system of USB, be used for the program of program code stored, user's voiceprint template and voice answer and vocal print database storing space module, secure memory space module, open storage space module, secure memory space module and open storage space module adopt the high-capacity FLASH storer to realize.

Images