IL119651A - Method and apparatus for supervised access to secured systems - Google Patents

Abstract

A method for establishing an identity of a speaker comprising the steps of: (a) providing a computerized system including at least two hardware installations employing at least two voice authentication algorithms, each of said at least two voice authentication algorithms being different from one another, each of said at least two hardware installations being for actuating one of said at least two voice authentication algorithms; and (b) independently analyzing a voice of the speaker by each of said at least two voice authentication algorithms for obtaining an independent positive or negative authentication of the voice by each of said at least two algorithms; wherein if every one of said at least two voice authentication algorithms provide a positive authentication, the speaker is positively identified, whereas, if at least one of said at least two voice authentication algorithms provides negative authentication, the speaker is negatively identified, whereas at least one of said at least two hardware installations is implemented in a secured- system, and at least another one of said at least two hardware installations is implemented in a securing-center, said at least one hardware installation implemented in said securing-center communicates with said at least one hardware installation implemented in said secured-system, such that all positive or negative identification data is established in said secured-system. 1079 י" ב בסיון התשס" ב - May 23, 2002

Description

METHOD AND APPARATUS FOR SUPERVISED ACCESS TO SECURED SYSTEMS and Apparatus for supervised access to secured systems This invention relates to method and devices for access to secured systems More particularly this invention relates to method and apparatus for improved access to secured systems using at least 2 independent voice authentication devices Background of the invention There are a number of applications with frequent access of users into the system with high security requirements Those systems are financial services such as stock trade confirmation bank account inquires and wire fund transfers Internet based electronic commerce computer networks homes and cars installations Currently available physical token authentication devices used for identifying individual such as crypto cards or limited access cards has a problem of low security protection Those cards can be lost stolen loaned to unauthorized persons or duplicated Another more sophisticated approach can be used to provide higher security protection using biometric authentication Biometric authentication is a means for positive identification by using unique body characteristics as an identifier such as fingerprints retinal scans facial recognition and voice pattern authentication Retinal scanning is based on a fact that retinal blood vessel patterns are unique and do not change over a person lifetime Although these feature provides high degree of security it suffers from expensive and complicated hardware and software Finger printing and facial recognition also requires expensive and complicated hardware and software Voice pattern authentication differs from voice pattern recognition Voice pattern recognition the speaker utters a word and the system determines the spoken word by selecting from a vocabulary Therefore voice recognition provides for the ability to recognize a spoken phrase not the identity of the speaker Voice verification also known as voice authentication speaker identity verification and voice print It is used to provide the speaker identification The techniques of voice verification has been described in US patents 083 These patents describes numerous methods for voice verification For example the speaker presumed identity may be verified using feature extraction and pattern matching algorithms These pattern matching is performed between the features of the digitized incoming voice print and those of the previously trained reference samples Voice authentication seeks to identify the speaker based solely on the spoken utterance An alternative techniques for voice identification are neural network comparison of the voice pattern using reference set password verification using selectively adjustable signal thresholds simultaneous voice recognition and verification The relatively low performance of voice verification technology has been one main reason for its cautious entry into the marketplace The Error Rate is a calculation algorithm of two parameters false acceptance and false rejection both varying on the degree of secure access required For example a typical error rate for false acceptance and rejection is about 2 By trying to set too low level of false rejection access denial the rate of false acceptance access grant will be too high Various techniques for security systems are described in following US patents 191 Those patents describe applications of various systems in different applications such as telephone networks computer networks cars and elevators These techniques and systems do not provide the required level of performance if low rate of false rejection is required then the rate of false acceptance will be unacceptably high and vice versa Therefore the need exists to provide more reliable and more secure system access with low false acceptance and rejection rates Brief description of the invention The system uses the technique of at least 2 independent voice recognition devices in order to provide the required level of acceptance and rejection rates Only coincidence of 2 channel acceptance will determine positive identification In order to eliminate the possibility of direct tampering with the voice channel devices the devices must be physically separated one at the supervising system security cente another at the device at the entrance to the secured system itself The method of 2 independent voice recognition devices using supervising system security center is the core of the present invention shows schematically how the incoming call from wire phone or wireless phone or computer phone is voice ied simultaneously by security center located in some remote site and secured system such as computer network according to prior Detailed Description of the Invention The process of the access to the secured system using supervising channel includes a number of steps Telephone call using a regular wire telephone or cellular wireless telephone or computer phone The telephone number of the secured system and the security center should be identical Therefore the call is received by 2 systems simultaneously In the call will be received by all devices to be secured by this security center since all this group will have identical telephone numbe These phones will be used only for receiving voice and will not have transmitter The block diagram of step 1 is shown in 2 Voice identification of the incoming call by the security center using specific algorithm of voice recognition for example using feature extraction and pattern matching and conventional verbal identification such as required system access password and social security The false rejection threshold should be preset to the low level say In this case false acceptance level will be in the order of After identification of the incoming call the security center will acknowledge the identification by transmitting audio pitch This audio pitch will be received both by caller and also routed to the specific secured block diagram of step 2 is shown in Voice identification of the incoming call by the secured system using algorithm different from the algorithm used by security center for example neural network processing The false rejection threshold should be preset to the low level say In this case false acceptance level will be in the order of The block diagram of step 3 is shown in Since security center and secured system are physically removed and since the process of identification in the security center will take some time interval ΔΤ the activation of the 2 simultaneous voice identification in the secured system will occur at ΔΤ after the receipt of the voice signal at the secured system This additional feature of preset time delay will ensure that no identification will occur before the acknowledgment from security center has been received The block diagram of step 4 is shown in 5 Only if both security center and secured system has simultaneous voice verification the caller has been identified and the process has been positively completed If one of the system does not verify the caller voice the process has not been positively completed and the access has been denied The combined rate of false acceptance will be much lower than each of the systems separately If false acceptance rate of each simultaneous voice identification channel is then combined rate of the false acceptance of the supervised secured system is or which is a very low number providing a very high security performance The block diagram of step 5 is shown in An alternative embodiment of the above system involves the use of RF transmitter and receivers in the network between the security center and secured systems is shown in In this embodiment the voice verification is performed not in parallel but in series and may be more lengthy than using parallel arrangement Description of the preferred embodiment The preferred embodiment of the aforementioned invention comprises a security center located at a different site from the secured system having following blocks a phone receiver and transmitter verbal interrogation device using human or computer operator and voice authentication device using voice verification algorithm The security center can be enclosed in a PC platform or used as system a secured device located at a site of a secured system having following blocks a phone receiver voice authentication device using voice verification algorithm different from the algorithm used in the security center and output interface providing secured system access in case of positive caller identification The secured device can be enclosed in a PC platform or used as system An alternative embodiment of the aforementioned invention comprises a security center located at a different site from the secured system having following blocks a phone receiver RF transmitter verbal interrogation device using human or computer operator and voice authentication device using voice verification algorithm The security center can be enclosed in a PC platform or used as system a secured device located at a site of a secured system having following blocks a RF receive voice authentication device using voice verification algorithm different from the algorithm used in the security center and output interface providing secured system access in case of positive caller identification The security center can be enclosed in a PC platform or used as system insufficientOCRQuality

Claims (22)

119651/2 5 Claims:

1. A method for establishing an identity of a speaker comprising the steps of: (a) providing a computerized system including at least two hardware installations employing at least two voice authentication algorithms, each of said at least two voice authentication algorithms being different from one another, each of said at least two hardware installations being for actuating one of said at least two voice authentication algorithms; and (b) independently analyzing a voice of the speaker by each of said at least two voice authentication algorithms for obtaining an independent positive or negative authentication of the voice by each of said at least two algorithms; wherein if every one of said at least two voice authentication algorithms provide a positive authentication, the speaker is positively identified, whereas, if at least one of said at least two voice authentication algorithms provides negative authentication, the speaker is negatively identified, whereas at least one of said at least two hardware installations is implemented in a secured-system, and at least another one of said at least two hardware installations is implemented in a securing-center, said at least one hardware installation implemented in said securing-center communicates with said at least one hardware installation implemented in said secured-system, such that all positive or negative identification data is established in said secured-system.

2. The method of claim 1, wherein each of said voice authentication algorithms is independently selected from the group consisting of text-dependent algorithms and text-independent algorithms. 11965-1/3

3. The method of claim 2, wherein each of said voice authentication algorithms is independently selected from the group consisting of feature extraction followed by pattern matching algorithms, neural network voice authentication algorithms, Dynamic Time Warping algorithm, Hidden Markov Model algorithm and vector quantization 1 algorithm.

4. The method of claim 1, wherein a false rejection threshold of each of said at least two algorithms is set to a level below 0.1%.

5. The method of claim 1, wherein the voice of the speaker is accepted for analysis by said computerized system via a remote communication mode.

6. The method of claim 5, wherein said remote communication mode is selected from the group consisting of wire telephone communication, cellular telephone communication, computer phone communication and radio communication.

7. The method of claim 1 , wherein said at least two hardware installations are remote from one another.

8. The method of claim 1, wherein said computerized system further includes a voice recognition algorithm for recognizing verbal data spoken by said speaker.

9. The method of claim 8, wherein said verbal data includes a spoken phrase selected from the group consisting of a name, an identification number, and a request. 7 119651/2

10. The method of claim 1, wherein said computerized system further includes a voice recognition algorithm for recognizing verbal data spoken by said speaker, the method further comprising the steps of positively or negatively recognizing said verbal data, and if said positive identity has been established, positively or negatively correlating between at least some of said verbal data and said speaker.

11. A method for establishing an identity of a speaker comprising the steps of: (a) providing a computerized system employing at least two voice authentication algorithms, each of said at least two voice authentication algorithms being different from one another; and (b) independently analyzing a voice of the speaker by each of said at least two voice authentication algorithms for obtaining an independent positive or negative authentication of the voice by each of said at least two algorithms; wherein if every one of said at least two voice authentication algorithms provide a positive authentication, the speaker is positively identified, whereas, if at least one of said at least two voice authentication algorithms provides negative authentication, the speaker is negatively identified, whereas said at least two voice authentication algorithms include a first voice authentication algorithm and a plurality of identical second voice authentication algorithms, said first voice authentication algorithm is implemented in a security-center, whereas said plurality of second voice authentication algorithms are implemented in a plurality of secured-systems, said security-center communicates with each of said plurality of secured-systems. δ 119651/3

12. A system for establishing an identity of a speaker comprising a computerized system including at least two hardware installations for actuating at least two voice authentication algorithms, each of said at least two hardware installations serves for actuating one of said at least two voice authentication algorithms, each of said at least two voice authentication algorithms being different from one another and serving for independently analyzing a voice of the speaker for obtaining an independent positive or negative authentication of the voice by each of said at least two algorithms, wherein if every one of said at least two voice authentication algorithms provide a positive authentication, the speaker is positively identified, whereas, if at least one of said at least two voice authentication algorithms provides negative authentication, the speaker is negatively identified, at least one of said at least two hardware installations is implemented in a secured-system, and at least another one of said at least two hardware installations is implemented in a securing-center, said at least one hardware installation implemented in said securing-center communicates with said at least one hardware installation implemented in said secured-system, such that all positive or negative identification data is established in said secured-system.

13. The method of claim 12, wherein each of said voice authentication algorithms is independently selected from the group consisting of text-dependent algorithms and text-independent algorithms.

14. The method of claim 13, wherein each of said voice authentication algorithms is independently selected from the group consisting of feature extraction followed by pattern matching algorithms, neural network voice authentication algorithms, Dynamic y 1 19651/3 Time Warping algorithm, Hidden Markov Model algorithm and vector quantization algorithm.

15. The system of claim 12, wherein a false rejection threshold of each of said at least two algorithms is set to a level below 0.1%.

16. The system of claim 12, wherein the voice of the speaker is accepted for analysis by said computerized system via a remote communication mode.

17. The system of claim 16, wherein said remote communication mode is selected from the group consisting of wire telephone communication, cellular telephone communication, computer phone communication and radio communication.

18. The system of claim 12, wherein said at least two hardware installations are remote from one another.

19. The system of claim 12, wherein said computerized system further includes a voice recognition algorithm for recognizing verbal data spoken by said speaker.

20. The system of claim 12, wherein said computerized system further includes a voice recognition algorithm for recognizing verbal data spoken by said speaker for positively or negatively recognizing said verbal data, and if said positive identity has been established, positively or negatively correlating between at least some of said verbal data and said speaker. 10 119651 3L

21. The system of claim 20, wherein said verbal data includes a spoken phrase selected from the group consisting of a name, an identification number, and a request.

22. A system for establishing an identity of a speaker comprising a computerized system including at least two voice authentication algorithms, each of said at least two voice authentication algorithms being different from one another and serving for independently analyzing a voice of the speaker for obtaining an independent positive or negative authentication of the voice by each of said at least two algorithms, wherein if every one of said at least two voice authentication algorithms provide a positive authentication, the speaker is positively identified, whereas, if at least one of said at least two voice authentication algorithms provides negative authentication, the speaker is negatively identified, said at least two voice authentication algorithms include a first voice authentication algorithm and a plurality of identical second voice authentication algorithms, said first voice authentication algorithm is implemented in a security-center, whereas said plurality of second voice authentication algorithms are implemented in a plurality of secured-systems, said security-center communicates with each of said plurality of secured-systems. . «ARK FRIEDMAN LT.D. PATENT ATTO R N EYS oAM U ELOFF B UI LDI NG ^ 7 HAOMAN I M STR EET 67897 TEL - AVIV I SRA EL