BIOMETRIC CONTROL SYSTEM AND METHOD

• US 20090187404A1
• Filed: 01/18/2008
• Published: 07/23/2009
• Est. Priority Date: 01/18/2008
• Status: Active Grant

• Current Assignee: Universidad de Chile
• Original Assignee: Universidad de Chile
• Inventors: YOMA, Nestor Jorge Becerra
• Granted Patent: US 8,050,920 B2
• Time in Patent Office: Days
• US Class Current: 704/246
• CPC Class Codes: G07C 9/33 by means of a password
• CPC Class Codes: G07C 9/37 using biometric data, e.g. fingerprints, iris scans or voice recognition
• CPC Class Codes: G10L 17/00 Speaker identification or verification techniques
• CPC Class Codes: G10L 17/22 Interactive procedures; Man-machine interfaces

Abstract

A large-scale attendance, productivity, activity and availability biometric control method using the telephone network, for individuals client users at their work places, with speaker verification technology based on limited enrolling data and short verification sentences, and the equipment associated to this method. Through the biometric control method by means of the individuals'"'"' identity (ID) verification using voice recognition, the individual is biometrically identified which allows for the registering of permanence, entrance or exit times. The type of performed activity, attendance at the work place, and which keeps records on the performed activity type. The method considers receiving and making intelligent calls to achieve an active control on individuals'"'"' activity.

Claims

1. A large-scale attendance, productivity, activity and availability biometric control method using the telephone network, for individual client users at their work places, with speaker verification technology based on limited enrolling data and short verification sentences, wherein said method comprises the steps of:

registering one or more system clients by entering, for each system client, a user name for a client user, an enrolling/verification sentence, a personal identification number (PIN) associated to the client user and a telephone number from where a subsequent enrolling and verification steps will take place;

enrolling the client user comprising requesting each client user to enter the associated PIN and to pronounce the enrolling/verification sentence, thereby providing voice signals;

generating a template and a reference average spectrum according to equation (1) with the voice signals provided by the user in the enrolling step; and

storing and indexing the template and the reference average spectrum with the PIN associated to the client user;

prompting the client user during a future verification event to enter the PIN and to pronounce the enrolling/verification sentence associated with the PIN, thereby providing a speech signal;

estimating a verification distance, in the future verification event, between the template indexed by the entered PIN and the speech signal pronounced by the client user;

using the reference average spectrum indexed by the entered PIN;

validating that the telephone number employed in the future verification event is the same as the telephone number employed in the enrolling step;

deciding if the speech signal pronounced by the client user has been validated based on the verification distance;

repetiting the steps of prompting, estimating, validating and deciding for a limited number of times, if the speech signal pronounced by the client user fails validation; and

accepting or rejecting the future verification event

2. The method according to claim 1, wherein in the enrolling step, the client user provides a plurality voice signals, wherein the plurality is “

N”

, and each of the “

N”

voice signals is processed by a digital analogical converter (A/D), and wherein the digital signals are parametrized, through the following procedure;

dividing each signal in a constant length “

windows”

sequence (based on the samples number);

performing an analysis on the spectral logarithm domain on each window, which results in a parameters vector, wherein each digitised signal is represented by a coefficients or parameters vectors sequence;

wherein O_n,j=[O_n,j,1, O_n,j,2, . . . , O_n,j,m, . . . , O_n,j,M] denotes the “

j”

frame or window vector “

M”

parameters, of signal “

n”

; and

, “

J_n”

is the “

n”

signal length expressed on window numbers;

thereby obtaining the user template and the reference average spectrum, wherein the reference average spectrum is represented by the average of coefficients within the enrolling speech signals and depends on;

the user, the enrolling/validation sentence, and the telephone line, and where the average spectrum corresponds to;

${\stackrel{\_}{O}}_e=\left[{\stackrel{\_}{O}}_{e,1},{\stackrel{\_}{O}}_{e,2},\mathrm{\dots <br><br>}<br><br>,{\stackrel{\_}{O}}_{e,m},\mathrm{\dots <br><br>}<br><br>,{\stackrel{\_}{O}}_{e,M}\right]<br><br><br><br>\mathrm{wherein},<br><br>{\stackrel{\_}{O}}_{e,m}=\frac{1}{N}<br><br>\underset{n=1}{\overset{N}{\sum <br><br>}}<br><br>{\stackrel{\_}{O}}_{n,m}<br><br><br><br>\mathrm{and}<br><br><br><br>{\stackrel{\_}{O}}_{n,m}=\frac{1}{J<br><br><br><br>n}<br><br>\underset{j=1}{\overset{J<br><br><br><br>n}{\sum <br><br>}}<br><br>{\stackrel{\_}{O}}_{n,j,m}.$

3. The method according to claim 2, wherein the received signal can be defined as:

y(t)=h(t)*[s(t)+r(t)]

where * denotes convolution;

r(t) additive noise; and

, h(t) o H(w), where H(w) is the Fourier transform of h(t), is the transmitter channel also denominated convolution noise.

4. The method according to claim 3, wherein the intra-speaker variability and additive noise compensation is performed with an unsupervised compensation of the signal, instead of adapting the user pattern or model, the intra-speaker variability compensation, consisting of eliminating the differences between the signal and the template or model if these are small.

5. The method according to claim 4 wherein O^E,C corresponds to the parametrized, equalized and compensated signal, and O_j^E,C=[O_j,1^E,C, O_j,2^E,C, . . . , O_j,m^E,C, . . . , O_j,M^E,C] is the “

M”

coefficient parameter vector corresponding to window “

j”

of O^E,C, O_jE,C=O_j^E+Δ

_j where Δ

_j is the correction vector applied to the O_j^E frame before compensating;

${\mathrm{\Delta <br><br>}}_j=\{\begin{array}{cc}\underset{X}{\arg }<br><br>\left[\mathrm{dist}<br><br>\left(O_j^E+X,P_{\mathrm{PIN}}\right)=0\right]& \mathrm{if}<br><br><br><br>\mathrm{dist}<br><br>\left(O_j^E,P_{\mathrm{PIN}}\right)\le <br><br>U<br><br><br><br>C\\ \underset{X}{\arg }<br><br>\left\{\mathrm{dist}<br><br>\left(O_j^E+X,P_{\mathrm{PIN}}\right)=f<br><br>\left[\mathrm{dist}<br><br>\left(O_j^E,P_{\mathrm{PIN}}\right)\right]\times <br><br>\mathrm{dist}<br><br>\left(O_j^E,P_{\mathrm{PIN}}\right)\right\}& \mathrm{if}<br><br><br><br>U<br><br><br><br>C\le <br><br>\mathrm{dist}\left(O_j^E,P_{\mathrm{PIN}}\right)\le <br><br>U<br><br><br><br>L\\ 0& \mathrm{if}<br><br><br><br>U<br><br><br><br>L\le <br><br>\mathrm{dist}<br><br>\left(O_j^E,P_{\mathrm{PIN}}\right)\end{array}$

wherein, dist(O_j^E, P_PIN) is the distance between window “

j”

, O_j^E, and the client model or template, P_PIN;

UC is the proximity threshold;

UL is the out-of-proximity threshold; and

, f[ ] is a function of dist(O_j^E, P_PIN).

6. Method according to claim 5, wherein the print voice pattern and the average spectrum are employed by the identity verification method to decide if the speech signal provided by the user, in a validation procedure by telephone, corresponds to the PIN-associated client.

7. The method according to claim 1, wherein the registration of a new client in the system, includes the steps of:

i) introducing the client data; and

ii) the enrolling procedure.

8. The method according to claim 1, wherein the use of this biometric control method is included in at least one of the following procedures:

i) the entrance or exit verification with an inbound call;

ii) the identity verification with an outbound call; and

iii) the reporting of activity.

9. The method according to claim 1, wherein the introduction of the client data includes:

full name;

the enrolling/verification sentence (e.g., first and family names);

a PIN associated to the client user in the system; and

, the telephone number allocated to the user from where enrolling and validation take place. Alternatively, further information related to the client could be entered, such as the national identification number.

10. The method according to claim 1, wherein the on-line enrolling takes place only once, after the client information is entered up in the system database, the enrolling procedure composing the steps of:

a) the user calling the system from the assigned telephone number, which is the same one that will be employed for identity validation;

b) the system requesting the user to enter his PIN;

c) the system requesting the user to pronounce the enrolling/verification sentence “

N”

times (2, 3, 4, etc.);

d) making use of the voice signals provided by the client, the system generating a template and the reference average spectrum associated to the client PIN, wherein the PIN identifies a given user from the others.

11. The method according to claim 10, wherein the system generates a PIN-associated-template or -model and a PIN-associated-average spectrum with the voice signals provided by the client user in the enrollment step.

12. The method according to claim 8, wherein the entrance or exit verification takes place every time the user requires, and composes:

a) the user calling the system from the telephone number that was assigned to him/her as described above;

b) the system requesting the user to enter his PIN;

c) the system requesting the user to pronounce once the short enrolling/validation sentence;

d) the system estimating the verification distance between the template or model indexed by the entered PIN and the speech signal pronounced by the user;

the verification distance estimation making use of the average spectrum indexed by the entered PIN; and

wherein the system can request the user to repeat the enrolling phrase to perform a more reliable decision;

e) the system registering the user entrance or exit with the following information;

date and hour of the validation;

the recorded voice signals used in the validation;

the voice identity validation result,

the result of the telephone number validation, and

f) the automatic voice identity validation decision can be validated itself at any time with the subjective analysis, by a referee, of the recorded speech signals.

13. The method according to claim 12, wherein the system compares the signal provided by the client user with the PIN-associated template by using the reference average spectrum in the identity validation as follows:

a) the identification distance, D_ID (shown in FIG. 8), that results from processing the verification speech signal with the template/model and the average of coefficients;

the estimation of D_ID employs an efficient intra-speaker variation and noise compensation method that allows short enrolling sessions;

this maximizes the usability of the system and makes possible the supervision of enrolling with human operators;

b) comparing D_ID with the acceptance or rejecting thresholds to take a decision about the user identity;

three possibilities can be considered;

accepted, doubtful, and rejected.

14. The method according to claim 1, wherein the report of type of activity takes place every time the user requires it and is composed of the following steps:

a) the user calling the system from the telephone number that was allocated to him/her as described above,

b) the system requesting the user to enter his/her PIN,

c) then the system prompting the user to pronounce once the enrolling/verification short sentence,

d) the system comparing the user signal with the template or model associated to his/her PIN and takes a decision about validating or not his/her identity, wherein the system can also identify the telephone number from where the call is made (by making use of “

ANI”

) and compare this telephone number with the one associated to the PIN, and the system can also request the user to repeat the enrolling/verification sentence to improve the reliability of the identity validation decision,

e) once the individual is validated, the system requiring the user to enter the type of activity he/she is starting, carrying out or finishing, wherein the type of activity can be entered up with touch tones or by voice if a speech recognizer is employed, and the informed activity could be a meeting, a training session, a customer'"'"'s visit, etc,

f) the user'"'"'s activity being registered with the following information;

date and hour;

recorded voice signals used in the validation;

result of the identity validation by voice;

type of activity.

result of the telephone number validation, and

g) the automatic voice identity validation decision being validated itself at any time with the subjective analysis, by a referee, of the recorded speech signals.

15. A speaker verification apparatus to control attendance, productivity and presence of individuals at work places, wherein said equipment performs said control as indicated on the method disclosed on claims 1 to 14, and where the speaker verification apparatus operates with limited enrolling data and short enrolling/verification sentences.