TW202450270A - A system, method and computer-readable medium thereof for interaction biometrics identity verification - Google Patents

Abstract

The present invention provides a system, method and computer-readable medium for interactive biometric authentication, including an interactive identification module, a facial recognition module, an interactive question-and-answer module, and a biometric recognition module. The invention presents a verification question to a user based on existing question-and-answer data, and uses the user's response to judge whether the answer is correct and whether the user is a live person. In this way, accurately confirms the correctness and authenticity of the answer video stream, and prevents individuals with malicious intent from using AI generated fake videos to pass identity verification, thereby enhancing the security of online services and reducing the cost of identity verification.

Description

An interactive liveness identification authentication system, method and computer-readable medium thereof

The present invention relates to an identity verification technology, and more particularly to an interactive liveness identification identity verification system, method, and computer-readable medium thereof.

With the rapid development of online services, the security of personal information on the Internet has become more important, so identity verification is very important.

The existing technology for biometric identity verification requires the user to shoot and register a video to be pre-stored in a biometric information database, where the video content is the image of the user reading out his account number, such as: 123456. Therefore, when verification is required, the user must face the audio and video communication device (such as a mobile phone) and read out the registered account number (such as 123456) so that the system can compare the user's face, voiceprint and content with the audio and video in the biometric information database to see if they match.

However, the existing technology must establish a user's living body information database to record the user's face, voiceprint and other biometric information, which not only costs a lot of money, but also causes inconvenience to the user in registration. Furthermore, in terms of information security, if the living body information database is leaked, the security problems it causes are even more immeasurable.

Therefore, how to provide a liveness identification authentication technology to authenticate users through liveness identification technology without the need to establish a new liveness information database, so as to significantly reduce manpower and material costs and improve information security.

To solve the above problems, the present invention provides an interactive live body recognition identity verification system, which includes: an interactive determination module, which, after receiving an identity verification request from a user device, issues at least one verification question to the user device and receives a video stream of a user's answer replied by the user device; an interactive question-and-answer module, which is communicatively connected to the interactive determination module to receive the video stream of the answer from the interactive determination module to obtain the user's verification answer in the video stream of the answer, and then the interactive question-and-answer module obtains a correct answer using a personal number to determine whether the user's verification answer is correct; and a live body recognition module, which is communicatively connected to the interactive determination module to receive the video stream of the answer from the interactive determination module. The interactive question-and-answer module determines whether the image and sound in the answer video stream are similar enough to achieve video and audio synchronization, and further determines whether the user in the answer video stream is alive. When the interactive question-and-answer module determines that the user's answer is correct and the liveness recognition module determines that the user is alive, the interactive recognition module confirms that the user has passed the identity recognition authentication, and transmits the identity recognition result of successful authentication to the user device; on the contrary, when the interactive question-and-answer module determines that the user's answer is incorrect, or the liveness recognition module determines that the user is not alive, the interactive recognition module confirms that the user has not passed the identity recognition authentication, and transmits the identity recognition result of failed authentication to the user device.

The present invention also provides an interactive live body identification identity verification method, which includes: after an interactive identification module receives an identity verification request from a user device, it sends at least one verification question to the user device and receives a video stream of a user's answer replied by the user device; an interactive question-and-answer module receives the video stream of the answer from the interactive identification module to obtain the user's verification answer in the video stream of the answer; after the interactive question-and-answer module obtains a correct answer using a personal number, it determines whether the user's verification answer is correct; and a live body identification module receives the video stream of the answer from the interactive identification module to determine whether the answer is correct. The interactive question-and-answer module determines whether the image and sound in the audio and video stream are similar enough to achieve audio and video synchronization, and then determines whether the user in the answer audio and video stream is alive. When the interactive question-and-answer module determines that the user's answer is correct and the liveness recognition module determines that the user is alive, the interactive determination module confirms that the user has passed the identity recognition authentication, and transmits the identity verification result of successful authentication to the user device; on the contrary, when the interactive question-and-answer module determines that the user's answer is incorrect, or the liveness recognition module determines that the user is not alive, the interactive determination module confirms that the user has not passed the identity recognition authentication, and transmits the identity verification result of failed authentication to the user device.

In the aforementioned embodiment, a face recognition module is further included, which is communicatively connected to the interactive determination module to receive the answer video stream from the interactive determination module and perform face recognition on the user in the answer video stream, wherein when the user passes the face recognition, the person number is obtained from a database to notify the interactive determination module that the user passes the face recognition.

In the aforementioned embodiment, the liveness recognition module captures the user's continuous facial images from the image and the user's continuous MFCC features from the voice to calculate the value of the complex cosine similarity between the continuous facial images and the continuous MFCC features, so as to obtain a synchronization similarity value from the value of the complex cosine similarity, and then determine whether the answer video stream is synchronized based on the synchronization similarity value.

In the aforementioned embodiment, the liveness recognition module further includes a liveness recognition model, which outputs a corresponding image vector group and a sound vector group in a multi-dimensional space after deep learning operation on the continuous face image and the continuous MFCC feature, and then calculates the value of the complex cosine similarity between the continuous face image and the continuous MFCC feature according to the image vector group and the sound vector group.

In the aforementioned embodiment, the interactive determination module determines whether there is a next verification question, and when there is a next verification question, the interactive question-and-answer module and the liveness recognition module determine whether the user in the other answer video stream replied by the user device has answered correctly and is alive.

As can be seen from the above, the interactive liveness identification authentication system, method and computer-readable medium of the present invention, after receiving the authentication request from the user device, asks the user a question to obtain an answer video stream, and performs facial recognition on the answer video stream to obtain the correct answer to confirm whether the user's authentication answer is correct, and at the same time determines whether the answer video stream is synchronized, thereby confirming whether the user in the answer video stream is alive.

Therefore, the present invention can accurately confirm the authenticity of the answer video stream by judging the correctness of the verification answer and the synchronization of the answer video stream, so as to prevent malicious people from using fake videos generated by artificial intelligence to pass identity authentication. It can also enable system service providers to directly use the question and answer data in the original database for identity authentication without spending more costs to establish a user's live database, thereby improving the security of online services and reducing the cost of identity authentication.

1: Interactive liveness identification authentication system

11: Interaction identification module

12: Face recognition module

13:Interactive question and answer module

14: Liveness recognition module

14a: Liveness recognition model

15: Database

9: User device (smartphone)

9a: Lens

S11 to S16, S21 to S211: Steps

Figure 1 is a schematic diagram of the architecture of the interactive liveness identification authentication system of the present invention.

Figure 1A is a schematic diagram of the video and audio synchronization recognition process of the present invention.

Figure 2 is a schematic diagram of the process of the interactive liveness identification authentication method of the present invention.

Figures 3A to 3C are schematic diagrams of a specific embodiment of the interactive liveness identification authentication system of the present invention.

The following is a specific and concrete example to illustrate the implementation of the present invention. People familiar with this technology can easily understand other advantages and effects of the present invention from the content disclosed in this manual.

It should be noted that the structures, proportions, sizes, etc. depicted in the drawings attached to this specification are only used to match the contents disclosed in the specification for understanding and reading by people familiar with this technology, and are not used to limit the restrictive conditions for the implementation of the present invention. Therefore, they have no substantial technical significance. Any modification of the structure, change of the proportion relationship or adjustment of the size should still fall within the scope of the technical content disclosed by the present invention without affecting the effects and purposes that can be achieved by the present invention. At the same time, the terms such as "one", "first", "second", "upper" and "lower" used in this specification are only used to facilitate the clarity of the description, and are not used to limit the scope of the implementation of the present invention. The changes or adjustments in their relative relationships shall be regarded as the scope of the implementation of the present invention without substantially changing the technical content.

FIG1 is a schematic diagram of the structure of the interactive liveness recognition identity verification system 1 of the present invention, which includes an interactive judgment module 11, a face recognition module 12, an interactive question-and-answer module 13, a liveness recognition module 14 having a liveness recognition model 14a, and a database 15.

Specifically, the interactive biometric authentication system 1 is established in a server (network server, general-purpose server, file server, storage unit server, etc.) or a computer or other electronic device with an appropriate computing mechanism, wherein each module in the interactive biometric authentication system 1 (the interactive judgment module 11, the face recognition module 12, the interactive The dynamic question-answering module 13 and the liveness recognition module 14 can be software, hardware or firmware; if it is hardware, it can be a processing unit, processor, computer or server with data processing and computing capabilities; if it is software or firmware, it can include instructions that can be executed by a processing unit, processor, computer or server, and can be installed on the same hardware device or distributed on different multiple hardware devices. In addition, the database 15 includes a hard disk (mechanical hard disk, solid state hard disk, etc.), and the database 15 can be a SQL (Structured Query Language) or NoSQL (Non-Structured Query Language) database, which is not limited here.

The interactive determination module 11 is communicatively (or electrically) connected to a user device 9, so that after receiving an identity verification request from the user device 9, the interactive determination module 11 sends at least one (including one or more) verification questions to a user through the user device 9 according to a plurality of verification questions preset in a database 15, and the user device 9 returns a video and audio stream of an answer replied by the user to the interactive determination module 11.

The face recognition module 12 is connected to the interactive judgment module 11 by communication (or electrical connection) to receive the answer video stream from the interactive judgment module 11, and the face recognition module 12 performs face recognition on the user in the answer video stream, and when the user passes the face recognition, obtains the corresponding personnel number of the user from a database 15 to provide it to the interactive judgment module 11 to notify the interactive judgment module 11 that the user passes the face recognition.

In one embodiment, the face recognition module 12 uses face recognition technology to obtain the user's facial feature data and compare it with the facial data in the database 15 to confirm the identity of the user, wherein the face recognition technology includes but is not limited to recognition algorithms based on facial features (feature-based recognition algorithms), recognition algorithms based on the entire face image (appearance-based recognition algorithms), template-based recognition algorithms (template-based recognition algorithms), recognition algorithms using neural networks (recognition algorithms using neural networks) and recognition algorithms using support vector machines (recognition algorithms using SVM), etc.

In one embodiment, the database 15 already has the correct answer required for general identity verification, and the correct answer is the answer entered by the user when registering an account.

The interactive question-and-answer module 13 is communicatively (or electrically) connected to the interactive determination module 11, so that after the interactive determination module 11 confirms that the user has passed the face recognition, the interactive determination module 11 provides the answer video stream to the interactive question-and-answer module 13, and the interactive question-and-answer module 13 uses the video recognition technology, voice recognition technology or lip recognition technology to obtain the user's verification answer in the answer video stream, and then the interactive question-and-answer module 13 obtains a correct answer from the database 15 according to the personnel number obtained by the face recognition module 12, so as to compare the user's verification answer with the correct answer to see if they are consistent, thereby judging whether the user's answer is correct.

In one embodiment, the speech recognition technology includes but is not limited to an automatic speech recognition tool (ASRT) based on a deep convolutional neural network or a long-short memory neural network; and the lip reading recognition technology includes but is not limited to LipNet based on a convolutional neural network (CNN), wherein the input of the LipNet architecture is a continuous lip image, which is calculated by a deep learning model based on a convolutional neural network (CNN) and then by CTC (Connectionist Temporal Classification) to obtain the output result of lip reading.

The liveness recognition module 14 is connected to the interactive judgment module 11 by communication (or electrical connection), so that after the interactive judgment module 11 confirms that the user has passed the face recognition, the interactive judgment module 11 provides the answer video stream to the liveness recognition module 14, and the liveness recognition module 14 executes a video and audio synchronization recognition process to identify the similarity between the image and sound in the answer video stream, and then judges whether the video and audio are synchronized according to the similarity between the sound and the image, thereby judging whether the user in the answer video stream is alive.

In one embodiment, as shown in FIG. 1A , the video and audio synchronization recognition process includes the following steps S11 to S16:

In step S11, the liveness recognition module 14 receives the answer video stream and divides the answer video stream into the image and sound files.

In step S12, after the liveness recognition module 14 obtains the image, the continuous facial image of the user is captured from the image.

In step S13, after the liveness recognition module 14 obtains the sound, it performs speech processing on the sound to extract MFCC (Mel-Frequency Cepstral Coefficient) features, wherein the liveness recognition module 14 processes the sound through pre-emphasis, frame capture, Hamming window, Fast Fourier Transform (FFT), Triangular Bandpass Filters, Log transform and Discrete Cosine Transform (DCT), thereby extracting the continuous MFCC features.

In step S14, the liveness recognition module 14 inputs the captured continuous facial image and the continuous MFCC features into a liveness recognition model 14a based on a neural network, so that the liveness recognition model 14a performs deep learning operations on the continuous facial image and the continuous MFCC features and outputs a corresponding image vector group and sound vector group in a multi-dimensional space, wherein the continuous facial image is converted into the image vector group through the liveness recognition model 14a, and the continuous MFCC features are converted into the sound vector group through the liveness recognition model 14a.

In step S15, the living body recognition module 14 uses a formula (1) to calculate the value of the complex cosine similarity (between -1 and 1) between the continuous face image and the continuous MFCC features according to the image vector group and the sound vector group, and obtains a synchronous similarity value according to the value of the complex cosine similarity by a statistical method, thereby calculating the distance in the dimensional space.

Among them, A refers to the image vector group converted by the continuous face image; and B refers to the sound vector group converted by the continuous MFCC features.

In one embodiment, the liveness recognition module 14 takes the median from the value of the multiple cosine similarity to obtain the synchronous similarity value; or, the liveness recognition module 14 uses an average, a weighted average, etc. to calculate the synchronous similarity value based on the value of the multiple cosine similarity.

For example, when the sampling rate is 100 Hz, 100 frames of MFCC features are taken per second, and the continuous MFCC features are 13-dimensional feature vectors. Then, 25 frames of continuous facial images are taken per second, and each frame of facial image corresponds to 4 frames of MFCC features. Furthermore, after the liveness recognition model 14a performs vector conversion, the liveness recognition module 14 calculates the values of the complex cosine similarities between the image vector group and the sound vector group one by one, for example: calculating the values of the cosine similarity between the "image vector of the facial image from the 1st to the 5th frame" and the "sound vector of the MFCC features from the 1st to the 20th frame", the "image vector of the facial image from the 2nd to the 6th frame" and the "sound vector of the MFCC features from the 2nd to the 6th frame". The value of the cosine similarity between the sound vector of the MFCC features of the 5th to 24th frames, the value of the cosine similarity between the image vector of the face image of the 3rd to 7th frames and the sound vector of the MFCC features of the 9th to 28th frames, and so on, is calculated to obtain the value of the complex cosine similarity, and then the synchronous similarity value is obtained by taking the median, average, weighted average and other statistical methods.

In step S16, the liveness recognition module 14 compares whether the synchronization similarity value is greater than a threshold value to determine whether the answer video stream has achieved video synchronization, thereby determining whether the user in the answer video stream is a live body, wherein if the synchronization similarity value is greater than the threshold value, the answer video stream is considered to be video synchronized; conversely, if the synchronization similarity value is less than the threshold value, the answer video stream is considered to be video asynchronous. For example: the threshold value of the synchronization similarity value can be set to 0.5, after calculation, the synchronization similarity value of sample A is 0.9, which is greater than 0.5, so sample A is video synchronized; and after calculation, the synchronization similarity value of sample B is 0.2, which is less than 0.5, so sample B is video asynchronous.

In one embodiment, when the interactive question-and-answer module 13 determines that the user's answer is correct and the liveness recognition module 14 determines that the user is alive, the interactive recognition module 11 confirms that the user has passed the identity recognition authentication and transmits the identity verification result (i.e., identity verification success) to the user device 9 to notify the user of the success of the identity verification through the user device 9; on the contrary, when the interactive question-and-answer module 13 determines that the user's answer is incorrect or the liveness recognition module 14 determines that the user is not alive, the interactive recognition module 11 confirms that the user has not passed the identity recognition authentication and transmits the identity verification result (i.e., identity verification failure) to the user device 9 to notify the user of the failure of the identity verification through the user device 9.

In one embodiment, the interactive determination module 11 determines whether there is a next verification question, wherein when there is a next verification question, the interactive liveness identification authentication system 1 determines whether the other answer video stream replied by the user is correct and the video and audio are synchronized.

FIG2 is a schematic diagram of the process of the interactive liveness identification authentication method of the present invention, and includes the following steps S21 to S211.

In step S21, a user uses a user device 9 to send an identity verification request to an interactive determination module 11.

In step S22, the interaction determination module 11 receives an identity verification request from the user device 9, and issues at least one verification question to a user through the user device 9 based on a plurality of questions preset in a database 15.

In step S23, the user accepts the verification question from the interactive determination module 11 through the user device 9, and the user uses the user device 9 to record an answer video stream, so that the user device 9 returns the answer video stream replied by the user to the interactive determination module 11.

In step S24, after the interaction determination module 11 receives the answer video stream from the user device 9, the answer video stream is transmitted to a face recognition module 12.

In step S25, the face recognition module 12 performs face recognition on the user in the answer video stream, and after the user passes the face recognition, the face recognition module 12 obtains the personnel number corresponding to the user from the database 15 and provides it to the interaction determination module 11, thereby notifying the interaction determination module 11 that the user passes the face recognition.

In step S26, the interaction determination module 11 confirms whether the user has passed the facial recognition. If the user has passed the facial recognition, step S27 is executed; otherwise, if the user has not passed the facial recognition, step S29 is executed.

In step S27, the interactive judgment module 11 provides the answer video stream to the interactive question-answer module 13 and the liveness recognition module 14, so that the interactive question-answer module 13 obtains the user's verification answer in the answer video stream, and obtains a correct answer from the database 15 according to the personnel number to determine whether the user's answer is correct, and the liveness recognition module 14 determines whether the video is synchronized according to the similarity between the sound and image in the answer video stream. If the user's answer is correct and the answer video stream is synchronized, step S28 is executed; otherwise, if the user's answer is incorrect or/and the answer video stream is not synchronized, step S29 is executed.

In step S28, the interaction determination module 11 determines whether there is a next verification question. If there is a next verification question, it determines whether the other answer video stream replied by the user is correct and the video and audio are synchronized; otherwise, if there is no next verification question, step S210 is executed.

In step S29, the interaction determination module 11 confirms that the user has not passed the identity authentication, and transmits the identity authentication result of the failed authentication to the user device 9.

In step S210, the interaction determination module 11 confirms that the user has passed the identity authentication, and transmits the successful identity authentication result to the user device 9.

In step S211, the user device 9 receives the identity verification result (including identity verification success or identity verification failure) from the interaction determination module 11 to notify the user.

In addition, the present invention also discloses a computer-readable medium, which is applied to a computing device or computer having a processor (e.g., CPU, GPU, etc.) and/or a memory, and stores instructions, and the computing device or computer can execute the computer-readable medium through the processor and/or memory to execute the above-mentioned method and each step when executing the computer-readable medium.

Figures 3A to 3C are schematic diagrams of a specific embodiment of the interactive liveness identification authentication system 1 of the present invention, and the similarities between this embodiment and the above embodiment will not be repeated.

In this embodiment, before a user wants to use service functions such as online customer service, the user first uses a smart phone 9 (i.e., a user device) to connect to an interactive biometric identity verification system 1 through a network to perform identity verification, wherein the smart phone 9 sends an identity verification request to the interactive determination module 11 in the interactive biometric identity verification system 1, and the interactive determination module 11 then returns at least one verification question to the smart phone 9. For example, as shown in FIG. 3A , the interactive determination module 11 returns the verification question “What elementary school do you attend?” to the smart phone 9, so that the verification question is displayed on the screen of the smart phone 9. At this time, as shown in FIG. 3B , the user turns on the camera 9a of the smart phone 9 to record the answer video stream of "Zhongli Elementary School", and the answer video stream includes the user's face.

Then, the smart phone 9 transmits the answer video stream back to the interactive determination module 11, and the interactive determination module 11 transmits the answer video stream to a face recognition module 12, so that the face recognition module 12 performs face recognition on the user in the answer video stream. After the user passes the face recognition, the face recognition module 12 obtains the corresponding personnel number of the user from a database 15, provides it to the interactive determination module 11, and notifies the user that the face recognition has been passed.

In this regard, the interactive determination module 11 provides the answer video stream to the interactive question-and-answer module 13 and the liveness recognition module 14, wherein the interactive question-and-answer module 13 obtains the user's verification answer (i.e. "Zhongli Elementary School") in the answer video stream, and then obtains a correct answer (i.e. "Zhongli Elementary School") from the database 15 according to the personnel number, whereby the interactive question-and-answer module 13 compares the verification answer and the correct answer to be consistent, and thus determines that the user's answer is correct.

Furthermore, the liveness recognition module 14 determines whether the video and audio are synchronized according to the similarity between the sound and the image in the answer video and audio stream. Specifically, after the liveness recognition module 14 divides the answer video and audio stream into the image and the sound files, it captures the user's continuous facial image from the image and the user's continuous MFCC features from the sound. The precision of the continuous facial image and the continuous MFCC features depends on the sampling rate, for example: 25 continuous facial images and 25 sets of continuous MFCC features (i.e., sound features) are taken per second, and each set of continuous MFCC features contains 4 MFCC features, so that each facial image and each set of MFCC features can be as precise as 40 milliseconds.

Afterwards, the liveness recognition module 14 calculates the value of the complex cosine similarity between the continuous face image and the continuous MFCC feature, and obtains a synchronization similarity value using a statistical method to determine whether the synchronization similarity value is greater than a threshold value. When the synchronization similarity value between the continuous face image and the continuous MFCC feature is less than a threshold value, it is determined that the answer video stream has achieved video synchronization, so the liveness recognition module 14 determines that the user is alive. At this time, the interactive judgment module 11 determines whether there is a next verification question, such as "What is the name of the first pet you raised?" If there is no next verification question, as shown in FIG3C, the interactive judgment module 11 confirms that the user has passed the identity authentication, and transmits the successful identity authentication result to the smart phone 9 to notify the user that the identity authentication is successful, so that the user can continue to use the subsequent online customer service and other services.

In summary, the interactive liveness identification authentication system, method and computer-readable medium of the present invention, after receiving the authentication request from the user device, asks the user a question to obtain an answer video stream, and obtains the correct answer after performing facial recognition on the answer video stream to confirm whether the user's authentication answer is correct, and at the same time determines whether the answer video stream is synchronized with the video, thereby confirming whether the user in the answer video stream is alive. Therefore, the present invention can accurately confirm the authenticity of the answer video stream by judging the correctness of the verification answer and the synchronization of the answer video stream, so as to prevent malicious people from using fake videos generated by artificial intelligence to pass identity authentication. It can also enable system service providers to directly use the question and answer data in the original database for identity authentication without spending more costs to establish a user's live database, thereby improving the security of online services and reducing the cost of identity authentication.

Furthermore, the interactive liveness identification authentication system, method and computer-readable medium of the present invention have at least the following technical differences and functions:

1. The present invention inputs a series of images (such as continuous face images) and sound features (continuous MFCC features) in audio and video into a living body recognition model based on neural networks to convert the images and sounds into vectors in multi-dimensional space. Then calculate the distance between the sound and the image in the dimensional space to determine whether the images are synchronized, thereby confirming whether the user is alive. Therefore, compared with the existing audio-visual synchronization judgment technology, it only targets Single characters of lip language and speech are recognized at corresponding time points, and the present invention can analyze the video and audio more precisely, greatly improving the accuracy of judgment.

2. The interactive liveness identification authentication system, method and computer-readable medium of the present invention can be widely used in various industries, such as communications, medical, finance, schools, and online sales industries, and the applied online service products are also rich, such as telecommunications customer service, online financial service system, online sales customer service system, and online medical treatment system, so as to quickly and conveniently provide a safer online authentication service environment.

The above implementation forms are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with this art may modify and change the above implementation forms without violating the spirit and scope of the present invention. Therefore, the scope of protection of the present invention should be as listed in the scope of the patent application.

1: Interactive liveness identification authentication system

11: Interaction identification module

12: Face recognition module

13:Interactive question and answer module

14: Liveness recognition module

14a: Liveness recognition model

15: Database

9: User device (smartphone)

Claims (11)

An interactive liveness identification authentication system comprises: An interactive identification module, after receiving an identity verification request from a user device, sends at least one verification question to the user device, and receives a video stream of a user's answer replied by the user device; An interactive question-and-answer module is communicatively connected to the interactive determination module to receive the answer video stream from the interactive determination module to obtain the user's verification answer in the answer video stream, and then the interactive question-and-answer module obtains a correct answer using a personnel number to determine whether the user's verification answer is correct; and A liveness recognition module is connected to the interactive identification module in communication to receive the answer video stream from the interactive identification module, so as to determine whether the similarity between the image and sound in the answer video stream reaches video synchronization, and further determine whether the user in the answer video stream is live. When the interactive question-and-answer module determines that the user's answer is correct and the liveness recognition module determines that the user is alive, the interactive recognition module confirms that the user has passed the identity recognition authentication, and transmits the identity verification result of successful authentication to the user device; on the contrary, when the interactive question-and-answer module determines that the user's answer is incorrect, or the liveness recognition module determines that the user is not alive, the interactive recognition module confirms that the user has not passed the identity recognition authentication, and transmits the identity verification result of failed authentication to the user device. The interactive liveness identification authentication system as described in claim 1 further includes a face recognition module, which is communicatively connected to the interactive determination module to receive the answer video stream from the interactive determination module, and perform face recognition on the user in the answer video stream, wherein when the user passes the face recognition, the person number is obtained from a database to notify the interactive determination module that the user passes the face recognition. The interactive liveness identification authentication system as described in claim 1, wherein the liveness identification module captures the user's continuous facial images from the image and the user's continuous MFCC features from the voice to calculate the value of the complex cosine similarity between the continuous facial images and the continuous MFCC features, so as to obtain a synchronization similarity value from the value of the complex cosine similarity, and then determine whether the answer video stream is synchronized based on the synchronization similarity value. The interactive liveness identification authentication system as described in claim 3, wherein the liveness identification module further includes a liveness identification model, which outputs the corresponding image vector group and sound vector group in the multi-dimensional space after deep learning operation on the continuous face image and the continuous MFCC feature, and then calculates the value of the complex cosine similarity between the continuous face image and the continuous MFCC feature according to the image vector group and the sound vector group. The interactive liveness identification authentication system as described in claim 1, wherein the interactive judgment module determines whether there is a next verification question, and when there is a next verification question, the interactive question-answering module and the liveness identification module determine whether the user in the other answer video stream replied by the user device has answered correctly and is alive. An interactive liveness identification authentication method includes: After receiving an identity verification request from a user device, an interactive identification module sends at least one verification question to the user device, and receives a video stream of a user's answer replied by the user device; An interactive question-and-answer module receives the answer video stream from the interactive judgment module to obtain the user's verification answer in the answer video stream; After the interactive question-and-answer module obtains a correct answer using a person ID, it determines whether the user's verification answer is correct; and A liveness recognition module receives the answer video stream from the interaction determination module to determine whether the similarity between the image and sound in the answer video stream reaches video and audio synchronization, and further determines whether the user in the answer video stream is alive. When the interactive question-and-answer module determines that the user's answer is correct and the liveness recognition module determines that the user is alive, the interactive recognition module confirms that the user has passed the identity recognition authentication, and transmits the identity verification result of successful authentication to the user device; on the contrary, when the interactive question-and-answer module determines that the user's answer is incorrect, or the liveness recognition module determines that the user is not alive, the interactive recognition module confirms that the user has not passed the identity recognition authentication, and transmits the identity verification result of failed authentication to the user device. The interactive liveness identification authentication method as described in claim 6 further includes receiving the answer video stream from the interactive determination module by a face recognition module to perform face recognition on the user in the answer video stream, wherein after the user passes the face recognition, the person number is obtained from a database to notify the interactive determination module that the user has passed the face recognition. The interactive liveness identification authentication method as described in claim 6 further includes the liveness identification module capturing the user's continuous facial images from the image and the user's continuous MFCC features from the voice to calculate the value of the complex cosine similarity between the continuous facial images and the continuous MFCC features, so as to obtain a synchronization similarity value from the value of the complex cosine similarity, and then determine whether the answer video stream is synchronized based on the synchronization similarity value. The interactive liveness identification authentication method as described in claim 8 further includes the liveness identification module performing deep learning operations on the continuous facial images and the continuous MFCC features, outputting corresponding image vector groups and sound vector groups in a multi-dimensional space, and then calculating the value of the complex cosine similarity between the continuous facial images and the continuous MFCC features based on the image vector group and the sound vector group. The interactive liveness identification authentication method as described in claim 6 further includes the interactive determination module determining whether there is a next verification question, and when there is a next verification question, the interactive question-and-answer module and the liveness identification module determine whether the user in the other answer video stream replied by the user device has answered correctly and is alive. A computer-readable medium, used in a computing device or a computer, stores instructions for executing an interactive liveness identification authentication method as described in any one of claims 6 to 10.

Images