Judgment device, video distribution system, judgment method, and program

Abstract

To determine whether or not a moving image in which a person is photographed is a synthetic moving image.SOLUTION: A determination device comprises: a moving image acquisition unit which acquires a moving image in which a person is photographed; a region extraction unit which extracts a region indicating a predetermined portion of the person included in the moving image; a pulse wave detection unit which detects a pulse wave from the region; and a moving image determination unit which determines whether or not the moving image is a synthetic moving image generated by a machine learning model on the basis of a detection result of the pulse wave.SELECTED DRAWING: Figure 3

Description

Technical Field

[0001] The present disclosure relates to a determination device, a video distribution system, a determination method, and a program.

Background Art

[0002] An information communication service that enables multiple users to share various electronic data using the Internet or the like is being used. In this type of information communication service, false information may be posted and spread due to user misunderstanding or malice.

[0003] For example, Patent Document 1 discloses an information determination device that determines whether false information is included in predetermined content by detecting a message including a predetermined negative word for the predetermined content from the posted message.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, in the prior art, it is not possible to determine whether false information is included in the posted electronic data itself. For example, in Patent Document 1, it is not possible to determine that false information is included in the predetermined content unless a message denying the predetermined content is detected.

[0006] One aspect of the present disclosure provides a technique for determining whether a video in which a person is photographed is a composite video.

Means for Solving the Problems

[0007] A determination device according to one aspect of the present disclosure includes a video acquisition unit that acquires a video of a person being filmed, a region extraction unit that extracts a region indicating a predetermined part of the person included in the video, a pulse wave detection unit that detects a pulse wave from the region, and a video determination unit that determines whether or not the video is a composite video generated by a machine learning model based on the pulse wave detection result. [Effects of the Invention]

[0008] According to one aspect of this disclosure, it is possible to determine whether or not a video in which a person is filmed is a composite video. [Brief explanation of the drawing]

[0009] [Figure 1] This is a block diagram showing an example of the overall configuration of a video streaming system. [Figure 2] A block diagram showing an example of a computer hardware configuration. [Figure 3] This block diagram shows an example of the functional configuration of a video distribution system in the first embodiment. [Figure 4] This is a block diagram showing an example of the functional configuration of the region extraction unit. [Figure 5] This is a block diagram showing an example of the functional configuration of the video analysis unit. [Figure 6] This is a sequence diagram showing an example of a video distribution method in the first embodiment. [Figure 7] This is a flowchart illustrating an example of region extraction processing. [Figure 8] This is a flowchart showing an example of a decision-making process. [Figure 9] This graph shows an example of a pulse wave signal. [Figure 10] This graph shows an example of the degree of fluctuation in pulse wave intervals. [Figure 11] This is a diagram showing an example of the basic scoring rules. [Figure 12] This is a diagram illustrating an example of the basic scoring rules. [Figure 13] This is a diagram illustrating an example of the basic scoring rules. [Figure 14] It is a diagram showing an example of the first correction rule. [Figure 15] It is a diagram showing an example of the second correction rule. [Figure 16] It is a block diagram showing an example of the functional configuration of the video distribution system in the second embodiment. [Figure 17] It is a sequence diagram showing an example of the video distribution method in the second embodiment.

Mode for Carrying Out the Invention

[0010] Hereinafter, each embodiment of the present disclosure will be described with reference to the accompanying drawings. In this specification and the drawings, components having substantially the same functional configuration are denoted by the same reference numerals, and redundant descriptions are omitted.

[0011] [First Embodiment] The first embodiment of the present disclosure is a video distribution system that provides a video distribution service. The video distribution service is an information communication service that distributes videos via an information communication network such as the Internet.

[0012] The video distribution service accepts video postings from a terminal device operated by a user and presents information about the posted videos to other users. In addition, the video distribution service accepts a video viewing request from a terminal device operated by another user and transmits the requested video to the terminal device. An example of the video distribution service is a video sharing service or a social networking service (SNS).

[0013] In an information communication service that enables users to share information with each other, false information different from the facts may be posted and spread. In particular, with the progress of video generation technology based on deep learning, the spread of fake videos (also called "deepfakes") that are difficult to distinguish from real videos has been increasing. For example, by using a generative adversarial network (GAN), it is possible to synthesize the face of another person onto a video in which a person is being filmed, creating a video as if another person was actually filmed. For example, an act of creating a video in which a socially influential person speaks false content and deliberately spreading it may become a problem.

[0014] Various methods for determining deepfakes have been proposed. For example, there is a method of analyzing, using a deep neural network, the light or color reflected in the pupils or the unnaturalness of the contour boundaries. However, since these deep neural networks have many layers and a high computational load, they are not widely available for general use.

[0015] The video distribution system in the present embodiment aims to determine whether a video in which a person is filmed is a synthetic video generated by a machine learning model. In particular, in the present embodiment, it aims to realize a technology that can determine a synthetic video even with few computational resources.

[0016] In one aspect, according to the present embodiment, it is possible to notify the user that the video being viewed is a synthetic video. By knowing that the video is a synthetic video, it becomes possible to suppress the spread of fake videos.

[0017] <Overall Configuration> The overall configuration of the video distribution system in the present embodiment will be described while referring to FIG. 1. FIG. 1 is a block diagram showing an example of the overall configuration of the video distribution system in the present embodiment.

[0018] As shown in Figure 1, the video distribution system 1 in this embodiment includes a video distribution device 10 and two terminal devices 20-1 and 20-2. The video distribution device 10 and the two terminal devices 20-1 and 20-2 are connected via a communication network N1 such as a LAN (Local Area Network) or the Internet, enabling data communication.

[0019] Note that the video distribution device 10 in this embodiment is an example of a determination device.

[0020] The video distribution device 10 is an information processing device such as a personal computer, workstation, or server that provides video distribution services. The video distribution device 10 receives a video posting request from terminal device 20-1 and stores the posted video in its storage device. The video distribution device 10 receives a video viewing request from terminal device 20-2 and transmits the requested video to terminal device 20-2.

[0021] Terminal device 20-1 is an information processing terminal such as a personal computer, tablet, or smartphone operated by a user (hereinafter also referred to as "poster") who posts videos to the video distribution service. Terminal device 20-1 transmits a video posting request to the video distribution device 10 in response to the poster's operation.

[0022] Terminal device 20-2 is an information processing terminal such as a personal computer, tablet terminal, or smartphone operated by a user (hereinafter also referred to as "viewer") who is viewing videos distributed from a video distribution service. In response to the viewer's operation, terminal device 20-2 sends a video viewing request to video distribution device 10. Terminal device 20-2 receives the video from video distribution device 10 and outputs it to a display device.

[0023] Note that the overall configuration of the video distribution system 1 shown in Figure 1 is just one example, and various system configurations are possible depending on the application and purpose. For example, one or more of the video distribution devices 10, terminal devices 20-1, and terminal devices 20-2 may be included in the video distribution system 1. For example, the video distribution device 10 may be implemented using multiple computers, or it may be implemented as a cloud computing service. The classification of devices such as the video distribution device 10 and terminal device 20 shown in Figure 1 is just one example.

[0024] <Hardware Configuration> The hardware configuration of the video distribution system 1 in this embodiment will be described with reference to Figure 2.

[0025] Computer Hardware Configuration In this embodiment, the video distribution device 10 and terminal device 20 are implemented, for example, by a computer. Figure 2 is a block diagram showing an example of the hardware configuration of the computer 500 in this embodiment.

[0026] As shown in Figure 2, the computer 500 includes a CPU (Central Processing Unit) 501, ROM (Read Only Memory) 502, RAM (Random Access Memory) 503, HDD (Hard Disk Drive) 504, input device 505, display device 506, communication interface 507, and external interface 508. The CPU 501, ROM 502, and RAM 503 form what is known as a computer. Each piece of hardware in the computer 500 is interconnected via a bus line 509. The input device 505 and display device 506 may also be used by connecting them to the external interface 508.

[0027] The CPU 501 is a processing unit that reads programs and data from a storage device such as the ROM 502 or HDD 504 onto the RAM 503 and executes processing, thereby realizing the overall control and functions of the computer 500. The computer 500 may have a GPU (Graphics Processing Unit) in addition to or instead of the CPU 501.

[0028] ROM502 is an example of non-volatile semiconductor memory (storage device) that can retain programs and data even when the power is turned off. ROM502 functions as the main memory, storing various programs and data necessary for the CPU501 to execute the programs installed on HDD504. Specifically, ROM502 stores boot programs such as BIOS (Basic Input / Output System) and EFI (Extensible Firmware Interface) that are executed when the computer 500 starts up, as well as OS (Operating System) settings, network settings, and other data.

[0029] RAM503 is an example of volatile semiconductor memory (storage device) whose programs and data are erased when the power is turned off. RAM503 includes, for example, DRAM (Dynamic Random Access Memory) and SRAM (Static Random Access Memory). RAM503 provides a working area that is expanded when various programs installed on HDD504 are executed by CPU501.

[0030] HDD504 is an example of a non-volatile storage device that stores programs and data. The programs and data stored in HDD504 include the operating system (OS), which is the basic software that controls the entire computer 500, and applications that provide various functions on the OS. Note that computer 500 may use a storage device that uses flash memory as its storage medium (e.g., SSD: Solid State Drive) instead of HDD504.

[0031] The input device 505 includes a touch panel used by the user to input various signals, operation keys and buttons, a keyboard and mouse, and a microphone for inputting sound data such as voice.

[0032] The display device 506 consists of a display such as a liquid crystal or organic EL (Electro-Luminescence) that displays a screen, and a speaker that outputs sound data such as audio.

[0033] Communication I / F 507 is an interface that connects to a communication network and allows computer 500 to perform data communication.

[0034] External I / F 508 is an interface for external devices. Examples of external devices include the drive device 510.

[0035] The drive device 510 is a device for setting the recording medium 511. The recording medium 511 here includes media that record information optically, electrically, or magnetically, such as CD-ROMs, flexible disks, and magneto-optical disks. The recording medium 511 may also include semiconductor memory that records information electrically, such as ROMs and flash memory. This allows the computer 500 to read and / or write to the recording medium 511 via the external I / F 508.

[0036] The various programs to be installed on the HDD 504 are installed, for example, when the distributed recording medium 511 is set in a drive device 510 connected to an external I / F 508, and the various programs recorded on the recording medium 511 are read by the drive device 510. Alternatively, the various programs to be installed on the HDD 504 may be downloaded via the communication I / F 507 from a network other than the communication network and installed that way.

[0037] <Functional Configuration> The functional configuration of the video distribution system 1 in this embodiment will be described with reference to Figure 3. Figure 3 is a block diagram showing an example of the functional configuration of the video distribution system 1 in this embodiment.

[0038] ≪Video Streaming Device≫ As shown in Figure 3, the video distribution device 10 in this embodiment includes a video storage unit 100, a video receiving unit 101, a request receiving unit 102, a video acquisition unit 103, a region extraction unit 104, a pulse wave detection unit 105, a video determination unit 106, and a video distribution unit 107.

[0039] The video storage unit 100 is implemented by the HDD 504 shown in Figure 2. The video storage unit 100 may also be implemented by an external storage device different from the video distribution device 10.

[0040] The video receiving unit 101, request receiving unit 102, video acquisition unit 103, region extraction unit 104, pulse wave detection unit 105, video determination unit 106, and video distribution unit 107 are realized by a process in which a program loaded from the HDD 504 shown in Figure 2 onto the RAM 503 is executed by the CPU 501.

[0041] The video storage unit 100 stores electronic data that indicates the content of the videos distributed by the video distribution service. Hereinafter, the electronic data indicating the content of the videos may simply be referred to as "videos." Each video stored in the video storage unit 100 is associated with identification information that identifies the video. The identification information may be, for example, a URL (Uniform Resource Locator) indicating the location of the video, or an identification number that the video distribution device 10 uniquely assigns to each video.

[0042] The video receiving unit 101 receives a video submission request from the terminal device 20-1. The submission request includes electronic data indicating the content of the video. The video receiving unit 101 generates identification information to identify the received video and stores it in the video storage unit 100 in association with the received video.

[0043] The request receiving unit 102 receives a video viewing request from the terminal device 20-2. The viewing request includes identification information that identifies the video specified by the viewer (hereinafter also referred to as the "target video").

[0044] The video acquisition unit 103 reads the target video from the video storage unit 100 based on the viewing request received by the request reception unit 102. In this embodiment, the target video is assumed to contain footage of at least one person.

[0045] The region extraction unit 104 extracts regions (hereinafter also referred to as "measurement regions") that indicate predetermined parts of a person included in the target video acquired by the video acquisition unit 103. In this embodiment, the region extraction unit 104 extracts multiple measurement regions that indicate different parts of the person. Specifically, the region extraction unit 104 extracts the areas where skin is exposed from among the right cheek, left cheek, forehead, and neck as measurement regions.

[0046] The pulse wave detection unit 105 detects pulse waves from each of the multiple measurement regions extracted by the region extraction unit 104. Specifically, the pulse wave detection unit 105 extracts the luminance change component of green light as a pulse wave signal in each measurement region.

[0047] In videos of living organisms, periodic changes in green light intensity are generated in the exposed skin areas due to hemoglobin changes synchronized with the heartbeat. Therefore, by extracting the green light intensity change component from the exposed skin areas of the video, it is possible to observe the pulse wave.

[0048] On the other hand, synthetic videos generated by machine learning models such as generative adversarial networks are created by extracting various scenes from videos similar to the original video and combining them. Therefore, such synthetic videos lack the continuity associated with changes in blood flow, and it is not possible to observe pulse waves. Consequently, it is possible to determine whether a target video is a synthetic video generated by a machine learning model based on the pulse waves detected from the video.

[0049] The video determination unit 106 determines whether the target video is a composite video based on the pulse wave detected by the pulse wave detection unit 105. Specifically, the video determination unit 106 calculates a score (hereinafter also called "truth value") indicating whether the target video is a composite video based on the pulse wave signals extracted from each measurement area. The video determination unit 106 may also calculate a truth value indicating whether the target video is a composite video by comparing the truth value with a predetermined threshold.

[0050] The video distribution unit 107 transmits the target video acquired by the video acquisition unit 103 and the determination result from the video determination unit 106 to the terminal device 20-2. The determination result transmitted by the video distribution unit 107 may be a truth value, a truth value indicating truth or falsehood, or may include both.

[0051] (Details of the region extraction unit) The functional configuration of the region extraction unit 104 will be explained in more detail with reference to Figure 4. Figure 4 is a diagram showing an example of the functional configuration of the region extraction unit.

[0052] As shown in Figure 4, the region extraction unit 104 in this embodiment takes a target video as input and outputs measurement region information indicating the measurement region. The region extraction unit 104 includes a face recognition unit 401, a skin determination unit 402, and a region determination unit 403.

[0053] The face recognition unit 401 recognizes a person's face from the target video. Next, the face recognition unit 401 identifies a predetermined area based on the face recognition result. In this embodiment, the predetermined area is the right cheek, left cheek, forehead, and neck.

[0054] The skin determination unit 402 determines whether or not skin is exposed in each region identified by the face recognition unit 401.

[0055] The region determination unit 403 identifies the region from the region identified by the face recognition unit 401 that the skin determination unit 402 has determined to have exposed skin. The region determination unit 403 determines the identified region to be the measurement region and outputs measurement region information indicating that measurement region. The measurement region information includes the range of the measurement region and the types of body parts included in the measurement region.

[0056] (Details of the video analysis unit) The functional configuration of the video detection unit 106 will be explained in more detail with reference to Figure 5. Figure 5 is a diagram showing an example of the functional configuration of the video detection unit.

[0057] As shown in Figure 5, the video determination unit 106 in this embodiment takes pulse wave signals detected in each measurement area as input and outputs a determination result regarding the target video. The video determination unit 106 includes an interval detection unit 601, a periodicity determination unit 602, a chaos analysis unit 603, an emotion determination unit 604, an authenticity calculation unit 605, an authenticity storage unit 606, and a result output unit 607.

[0058] The interval detection unit 601 calculates the interval between pulse waves for the pulse wave signals detected in each measurement area. The interval detection unit 601 generates time-series data of pulse wave intervals, which are obtained by arranging the pulse wave intervals in a time series.

[0059] The periodicity determination unit 602 determines whether the pulse wave has a predetermined periodicity for the pulse wave signal detected in each measurement area. If the pulse wave does not have the predetermined periodicity, the periodicity determination unit 602 sets the pulse wave detection result in that measurement area to "not detected".

[0060] The chaos analysis unit 603 analyzes the fluctuations in pulse wave intervals for the pulse wave signals detected in each measurement area. Specifically, the chaos analysis unit 603 calculates the degree of fluctuation in pulse wave intervals based on the time-series data of pulse wave intervals generated by the interval detection unit 601.

[0061] The emotion determination unit 604 determines the emotion based on the degree of fluctuation of the pulse wave interval calculated by the chaos analysis unit 603. Specifically, the emotion determination unit 604 classifies the emotion into one of several predetermined emotions by comparing the degree of fluctuation of the pulse wave interval with a predetermined threshold.

[0062] The truthfulness calculation unit 605 calculates the truthfulness based on the pulse wave detection result and the emotion determination result. The truthfulness calculation unit 605 calculates the truthfulness at predetermined time intervals (for example, about 30 seconds).

[0063] In this embodiment, the truth value takes a value between 0 and 100. A higher truth value indicates a higher probability that the video is real (in other words, not a composite video), while a lower truth value indicates a higher probability that the video is false (in other words, a composite video).

[0064] The truthfulness storage unit 606 stores the truthfulness calculated by the truthfulness calculation unit 605 for each target video in a time series. Therefore, the truthfulness storage unit 606 stores time series data of the truthfulness for each target video.

[0065] The result output unit 607 calculates the average truth value for the target video based on the time-series truth value data read from the truth value storage unit 606. The result output unit 607 generates a judgment result for the target video based on the average truth value.

[0066] ≪Terminal Devices≫ As shown in Figure 3, terminal device 20-1 in this embodiment includes a video transmission unit 201. Terminal device 20-2 in this embodiment includes a video request unit 202 and a video display unit 203.

[0067] Note that terminal devices 20-1 and 20-2 may have the same functional configuration. That is, terminal device 20-1 may further include a video request unit 202 and a video display unit 203. Also, terminal device 20-2 may further include a video transmission unit 201.

[0068] The video transmission unit 201, video request unit 202, and video display unit 203 are realized by a process in which a program loaded from the HDD 504 shown in Figure 2 onto the RAM 503 is executed by the CPU 501.

[0069] The video transmission unit 201 sends a video submission request to the video distribution device 10 in response to the uploader's operation. The video submission request includes electronic data indicating the content of the video specified by the uploader.

[0070] The video request unit 202 transmits a video viewing request to the video distribution device 10 in response to the viewer's operation. The video viewing request includes identification information that identifies the video specified by the viewer.

[0071] The video display unit 203 receives the target video and the judgment result related to the target video from the video distribution device 10. The video display unit 203 outputs the target video and the judgment result to the display device 506.

[0072] <Processing Procedure> The processing procedure of the video distribution method executed by the video distribution system 1 in this embodiment will be described with reference to Figure 6. Figure 6 is a sequence diagram showing an example of the processing procedure of the video distribution method in this embodiment.

[0073] In step S1, the video transmission unit 201 of the terminal device 20-1 accepts a video posting operation from the poster. The video posting operation is performed, for example, by specifying the video to be posted on the video posting screen presented by the video distribution system 1.

[0074] Next, the video transmission unit 201 sends a video submission request to the video distribution device 10. The video submission request includes electronic data indicating the content of the video specified by the submitter.

[0075] In step S2, the video receiving unit 101 of the video distribution device 10 receives a video posting request from the terminal device 20-1. Next, the video receiving unit 101 acquires the video included in the posting request. Subsequently, the video receiving unit 101 generates identification information to identify the acquired video. Then, the video receiving unit 101 associates the video with the identification information and stores it in the video storage unit 100.

[0076] In step S3, the video request unit 202 of the terminal device 20-2 accepts a video viewing operation from the viewer. The video viewing operation is, for example, an operation in which the viewer specifies the target video they wish to view on the video list screen presented by the video distribution system 1.

[0077] Next, the video request unit 202 sends a video viewing request to the video distribution device 10. The video viewing request includes identification information that identifies the target video specified by the viewer.

[0078] In step S4, the request receiving unit 102 of the video distribution device 10 receives a video viewing request from the terminal device 20-2. Next, the request receiving unit 102 sends the identification information included in the video viewing request to the video acquisition unit 103.

[0079] The video acquisition unit 103 receives identification information from the request receiving unit 102. Next, the video acquisition unit 103 identifies the target video stored in the video storage unit 100 based on the identification information. Subsequently, the video acquisition unit 103 reads the target video from the video storage unit 100. Then, the video acquisition unit 103 sends the read target video to the area extraction unit 104 and the video distribution unit 107.

[0080] In step S5, the region extraction unit 104 of the video distribution device 10 receives the target video from the video acquisition unit 103. Next, the region extraction unit 104 extracts the measurement region from the target video. Then, the region extraction unit 104 sends the measurement region information indicating the measurement region to the pulse wave detection unit 105.

[0081] <<Region Extraction Process>> The region extraction process in this embodiment (step S5 in Figure 6) will be described in more detail with reference to Figure 7. Figure 7 is a flowchart showing an example of the region extraction process in this embodiment.

[0082] In step S11, the face recognition unit 401 of the region extraction unit 104 recognizes a person's face from the target video. Specifically, the face recognition unit 401 first analyzes the facial features (face range, orientation, feature points of each part, etc.) from the target video. For the analysis of facial features, for example, a contour detection algorithm and a feature point extraction algorithm can be used.

[0083] Next, the face recognition unit 401 identifies a predetermined area based on the analysis results of the facial features. In this embodiment, the predetermined areas are the right cheek, left cheek, forehead, and neck. The face recognition unit 401 then sends area information indicating the predetermined area to the skin determination unit 402.

[0084] In step S12, the skin determination unit 402 of the region extraction unit 104 receives region information from the face recognition unit 401. Next, the skin determination unit 402 determines whether or not skin is exposed for each region indicated in the region information. For example, the skin determination unit 402 determines whether or not the color representing each region (e.g., the average color) falls within the range of human skin color. Then, the skin determination unit 402 sends skin region information indicating the regions where skin is exposed to the region determination unit 403.

[0085] In step S13, the region determination unit 403 of the region extraction unit 104 receives skin region information from the skin determination unit 402. Next, the region determination unit 403 determines the region indicated in the skin region information as the measurement region. Then, the region determination unit 403 outputs measurement region information indicating the measurement region.

[0086] Let's return to Figure 6 for explanation. In step S6, the pulse wave detection unit 105 of the video distribution device 10 receives measurement area information from the area extraction unit 104. Next, the pulse wave detection unit 105 identifies the measurement area in the target video based on the measurement area information.

[0087] Next, the pulse wave detection unit 105 generates a time-series signal of skin color in each identified measurement area. Then, the pulse wave detection unit 105 extracts the luminance change component of green light from the time-series signal of skin color. The pulse wave detection unit 105 also performs noise reduction on the extracted luminance change component of green light using a square wave correlation filter. Finally, the pulse wave detection unit 105 sends the luminance change component after noise reduction as a pulse wave signal to the video determination unit 106.

[0088] In step S7, the video determination unit 106 of the video distribution device 10 receives pulse wave signals from the pulse wave detection unit 105 in each measurement area. Next, the video determination unit 106 determines whether or not the target video is a composite video based on each pulse wave signal. Then, the video determination unit 106 sends the determination result regarding the target video to the video distribution unit 107.

[0089] ≪Decision Processing≫ The determination process in this embodiment (step S7 in Figure 6) will be described in more detail with reference to Figure 8. Figure 8 is a flowchart showing an example of the determination process in this embodiment.

[0090] In step S21, the interval detection unit 601 of the video determination unit 106 detects the peak points of the pulse wave signals in each measurement area. Next, the interval detection unit 601 calculates the interval between each detected peak point. Subsequently, the interval detection unit 601 generates time-series data of pulse wave intervals by arranging the intervals between each peak point in time series. Then, the interval detection unit 601 sends the time-series data of pulse wave intervals to the periodicity determination unit 602 and the chaos analysis unit 603.

[0091] Figure 9 is a graph showing an example of a pulse wave signal. The graph in Figure 9 shows the pulse wave signal PW, with the horizontal axis t representing time (milliseconds) and the vertical axis A representing the amplitude strength of the pulse wave. As shown in Figure 9, the pulse wave signal PW is triangular in shape, reflecting the fluctuations in blood flow due to the heartbeat. The interval detection unit 601 detects the peak points P1 to Pn with the strongest blood flow from the pulse wave signal PW and calculates the pulse wave intervals d1 to dn.

[0092] Let's return to Figure 8 for explanation. In step S22, the periodicity determination unit 602 of the video determination unit 106 receives time-series data of pulse wave intervals from the interval detection unit 601. Next, the periodicity determination unit 602 determines whether or not the pulse wave has periodicity based on the time-series data of pulse wave intervals in each measurement area. If it is determined that the pulse wave does not have periodicity, the periodicity determination unit 602 sets the pulse wave detection result in that measurement area to "not detected".

[0093] Furthermore, the periodicity determination unit 602 determines whether the period of the pulse wave in a measurement area where periodicity has been determined to exist is within the range that a human heart rate can take. Specifically, the periodicity determination unit 602 first calculates the number of peak points P in a predetermined time unit. Next, the periodicity determination unit 602 determines whether the number of peak points P is within the range that a human heart rate can take. The range that a human heart rate can take during non-exercise can be defined as, for example, about 40 to 100 beats per minute. If the number of peak points P is outside the range that a human heart rate can take, the periodicity determination unit 602 sets the pulse wave detection result in that measurement area to "not detected".

[0094] The periodicity determination unit 602 determines that the pulse wave has a predetermined periodicity in a measurement area and sets the pulse wave detection result in that measurement area as "detected". The periodicity determination unit 602 sends the pulse wave detection results for each measurement area to the truth value calculation unit 605.

[0095] In step S23, the chaos analysis unit 603 of the video determination unit 106 receives time-series data of pulse wave intervals from the interval detection unit 601. Next, the chaos analysis unit 603 calculates the degree of fluctuation of the pulse wave intervals based on the time-series data of the pulse wave intervals. In this embodiment, the degree of fluctuation of the pulse wave intervals is the maximum Lyapunov exponent. The chaos analysis unit 603 then sends the maximum Lyapunov exponent for each measurement region to the emotion determination unit 604.

[0096] Figure 10 is a graph showing an example of the degree of fluctuation in pulse wave interval. The graph shown in Figure 10 is also called a Lorentz plot. A Lorentz plot plots time-series data of pulse intervals on the coordinate (dn, dn-1) for n=1, 2, ... with the horizontal axis representing the pulse wave interval dn and the vertical axis representing the pulse wave interval dn-1 (both in milliseconds).

[0097] The maximum Lyapunov exponent can be calculated using the coordinates (dn, dn-1) in the Lorentz plot, for example, by equation (1).

[0098]

number

[0099] However, λ is the maximum Lyapunov exponent, M is the total sample time, and d is the pattern distance between time k and time k-1 in the time series data (distance on the 2D plane in the Lorentz plot).

[0100] Let's return to Figure 8 for explanation. In step S24, the emotion determination unit 604 of the video determination unit 106 receives the degree of fluctuation of the pulse wave interval in each measurement area from the chaos analysis unit 603. Next, the emotion determination unit 604 determines the emotion based on the degree of fluctuation of the pulse wave interval in each measurement area. The emotion determination unit 604 classifies the emotion into one of several predetermined emotions by, for example, comparing the degree of fluctuation of the pulse wave interval with a predetermined threshold. The emotion determination unit 604 sends the result of the emotion determination in each measurement area to the truth value calculation unit 605.

[0101] For example, the emotion determination unit 604 may determine that a "negative emotion" is present, such as brain fatigue, anxiety, or depression, if the maximum Lyapunov exponent is below a predetermined negative threshold (for example, around -0.6). Alternatively, for example, the emotion determination unit 604 may determine that a "positive emotion" is present, such as brain fatigue, anxiety, or depression, if the maximum Lyapunov exponent is 0 or greater.

[0102] The types and number of emotions that the emotion determination unit 604 determines, as well as the thresholds for determining them, can be arbitrarily defined. For example, the emotion determination unit 604 may determine four emotions: "stress-free," "active," "slightly fatigued," and "fatigued."

[0103] In step S25, the authenticity calculation unit 605 of the video determination unit 106 receives the pulse wave detection results for each measurement area from the periodicity determination unit 602. The authenticity calculation unit 605 also receives the emotion determination results for each measurement area from the emotion determination unit 604.

[0104] Next, the truthfulness calculation unit 605 determines the basic truthfulness score based on the pulse wave detection result in the cheek. The truthfulness calculation unit 605 determines the basic truthfulness score according to predetermined basic scoring rules.

[0105] Figure 11 shows an example of the basic scoring rules. As shown in Figure 11, the basic scoring rules determine the degree of truthfulness by applying conditions for the presence or absence of pulse wave detection in the cheek, depending on the orientation of the face captured in the target video.

[0106] For example, if the target video shows a face facing forward, and pulse waves are detected in both cheeks, the basic accuracy score will be 90 points. On the other hand, if pulse waves are detected in only one cheek, the basic accuracy score will be 80 points. Furthermore, if no pulse waves are detected in either cheek, the basic accuracy score will be 20 points.

[0107] For example, if the target video shows a face in profile, and a pulse wave is detected on the cheek on the side being filmed, the basic accuracy score will be 80 points. On the other hand, if no pulse wave is detected on the cheek on the side being filmed, the basic accuracy score will be 50 points.

[0108] Figures 12 and 13 illustrate an example of the basic scoring rules. Figure 12 shows an example of a target video 410 in which a face is facing forward. As shown in Figure 12, in a target video 410 in which a face is facing forward, the skin is often exposed in the right cheek region 411 and the left cheek region 412. On the other hand, the forehead region 413 may not be exposed due to a hat or bangs, etc. Also, the neck region 414 may not be exposed due to clothing, etc. Therefore, in a target video 410 in which a face is facing forward, the basic truthfulness is determined based on the pulse wave detection results in both cheeks.

[0109] If no pulse waves are detected in either cheek, it is highly likely that the person's face was synthesized using a machine learning model. Therefore, it is advisable to set the base authenticity value low. On the other hand, if pulse waves are detected in both cheeks, it is highly likely that the person's face was actually captured in a photograph. Therefore, it is advisable to set the base authenticity value high. Note that since machine learning models often synthesize the entire face in synthesized videos, a high base authenticity value can also be set if pulse waves are detected in only one cheek.

[0110] Figure 13 shows an example of a target video 420 in which a face is viewed from the side. As shown in Figure 13, in target videos 420 in which a face is viewed from the side, the skin is often exposed on only one cheek (the left cheek region 412 in the example in Figure 13). Therefore, in target videos 420 in which a face is viewed from the side, the basic truth value is determined based on the pulse wave detection result on one cheek.

[0111] If no pulse wave is detected on one cheek, it is difficult to determine whether the person's face was synthesized using a machine learning model. Therefore, it is best to set the baseline authenticity level to a medium. On the other hand, if a pulse wave is detected on one cheek, there is a high probability that the person's face was actually captured. Therefore, it is best to set the baseline authenticity level to a high level.

[0112] Let's return to Figure 8 for explanation. In step S26, the truthfulness calculation unit 605 of the video determination unit 106 corrects the basic truthfulness determined in step S25 based on the pulse wave detection result in the forehead or neck. The truthfulness calculation unit 605 corrects the truthfulness according to a predetermined first correction rule. The first correction rule is a rule for correcting the truthfulness based on the pulse wave detection result.

[0113] Figure 14 shows an example of the first correction rule. As shown in Figure 14, the first correction rule applies conditions to whether or not pulse waves are measured and detected on the forehead and neck, and adds or subtracts from the truth value. For example, if a pulse wave is detected on the forehead, the truth value should be added (e.g., +10 points). Conversely, if a pulse wave is not detected on the forehead, the truth value should be subtracted (e.g., -10 points). Since synthesized videos created by machine learning models often synthesize the entire face, if a pulse wave can be detected on the forehead, there is a high probability that the person's face was actually captured.

[0114] Furthermore, for example, if no pulse wave is detected in the neck, it is advisable to subtract from the truthfulness score (e.g., -10 points). On the other hand, if a pulse wave is detected in the neck, no addition or subtraction of the truthfulness score is necessary. Since machine learning models often synthesize only faces in synthesized videos, even if a pulse wave is detected in the neck, it is highly likely that it was captured in the video before synthesis.

[0115] Furthermore, if pulse waves cannot be measured in the forehead and neck, it is not necessary to adjust the accuracy of the readings. This is because it is common for measurements to be impossible in these areas due to clothing or hairstyle.

[0116] Let's return to Figure 8 for explanation. In step S27, the truthfulness calculation unit 605 of the video judgment unit 106 further corrects the basic truthfulness (or the truthfulness corrected in step S26) determined in step S25 based on the emotion judgment result. The truthfulness calculation unit 605 corrects the truthfulness according to a predetermined second correction rule. The second correction rule is a rule for correcting the truthfulness based on the emotion judgment result.

[0117] Figure 15 shows an example of the second correction rule. As shown in Figure 15, the second correction rule applies conditions to the emotion judgment result and adds or subtracts from the truth value. For example, if the emotion judged on the cheek is different from the emotion judged on the neck, it is good to subtract a large amount from the truth value (e.g., -100 points). Since synthesized videos created by machine learning models often only synthesize the face, if the emotion differs between the cheek, which is within the face region, and the neck, which is outside the face region, it is highly likely that the video was synthesized by a machine learning model.

[0118] Furthermore, if, for example, no emotional fluctuations are observed after multiple attempts to assess emotions within the same measurement area, it is advisable to deduct points from the truthfulness score (e.g., -20 points). Since the absence of emotional fluctuations over a long period of time is unlikely to occur unless the person has undergone special training, it is reasonable to conclude that the video is likely a synthesized video created by a machine learning model.

[0119] Let's return to Figure 8 for explanation. The truthfulness calculation unit 605 stores the truthfulness calculated in steps S25 to S27 in the truthfulness storage unit 606. The truthfulness storage unit 606 stores the truthfulness for each target video in chronological order.

[0120] In step S28, the video determination unit 106 determines whether the target video has finished or not. If the target video has finished (YES), the video determination unit 106 proceeds to step S29. On the other hand, if the target video has not finished (NO), the video determination unit 106 returns to step S21.

[0121] Subsequently, the video determination unit 106 repeats the process from step S21 to step S27 until the target video ends. As a result, time-series data of the truthfulness of the entire target video is stored in the truthfulness storage unit 606.

[0122] In step S29, the result output unit 607 of the video judgment unit 106 reads time-series data of the truthfulness of the target video from the truthfulness storage unit 606. Next, the result output unit 607 calculates the average truthfulness of the target video. Then, the result output unit 607 outputs a judgment result for the target video based on the average truthfulness.

[0123] The result of the judgment may be the average value of the truthfulness score. The result of the judgment may also be a truth value indicating whether the average value of the truthfulness score is greater than or equal to a predetermined threshold (e.g., 50). For example, if the average value of the truthfulness score is greater than or equal to the threshold, it will be a true value (e.g., 1) indicating that the target video is true, and if the average value of the truthfulness score is less than the threshold, it will be a false value (e.g., 0) indicating that the target video is false. The result of the judgment may include both the average value of the truthfulness score and a truth value indicating truth or falsehood.

[0124] Let's return to Figure 6 for explanation. In step S8, the video distribution unit 107 of the video distribution device 10 receives the target video from the video acquisition unit 103. The video distribution unit 107 also receives the judgment result regarding the target video from the video judgment unit 106. Then, the video distribution unit 107 transmits the target video and the judgment result to the terminal device 20-2.

[0125] In step S9, the video display unit 203 of the terminal device 20-2 receives the target video and the judgment result related to the target video from the video distribution device 10. The video display unit 203 then outputs the target video and the judgment result to the display device 506.

[0126] The viewer views the target video displayed on the display device 506. At this time, the display device 506 displays the target video along with the judgment result regarding the target video. If the judgment result indicates that it is a composite video, the viewer can know that there is a high possibility that the video they are viewing is a composite video.

[0127] Viewers who discover that a video is a composite can post a message or comment indicating that the video is fake to the video distribution system 1. Alternatively, viewers who discover that a video is a composite can refrain from further spreading the video, which is highly likely to be a composite. As a result, the spread of composite videos on the video distribution system 1 can be suppressed.

[0128] <Effects of the First Embodiment> In this embodiment, the video distribution device 10 detects pulse waves from a region indicating a predetermined part of a person included in the video, and determines whether the video is a composite video generated by a machine learning model based on the pulse wave detection result. Composite videos generated by machine learning models are created by combining various scenes extracted from other videos. Therefore, such composite videos lack continuity associated with changes in blood flow, making it impossible to detect pulse waves. Accordingly, this embodiment makes it possible to determine whether a video in which a person is filmed is a composite video.

[0129] In this embodiment, the video distribution device 10 may determine that a video is a composite video when no pulse wave is detected from a region indicating a predetermined part of a person. The predetermined part may be the cheek. In composite videos generated by machine learning models, a composite video may be created in which the face of another person is superimposed on a video in which a person is being filmed. The cheek is a part of a person's face where the skin is often exposed. Therefore, according to this embodiment, it is possible to accurately determine whether a video has a person's face superimposed on it.

[0130] In this embodiment, the video distribution device 10 may extract multiple regions representing different parts of a person and calculate a score indicating whether or not the video is a composite video based on the pulse wave detection results in each region. By making a comprehensive determination using the pulse wave detection results for each of the multiple regions, it becomes possible to determine various composite videos in which parts of a person's face have been composited. Therefore, according to this embodiment, it is possible to accurately determine videos in which parts of a person's face have been composited.

[0131] In this embodiment, the video distribution device 10 may correct the score based on the pulse wave detection result in the person's cheek using the pulse wave detection result in the person's forehead or neck. The cheek is a part of the face where skin is often exposed. The forehead and neck may not have exposed skin. By using the pulse wave detection result in the forehead or neck in addition to the pulse wave detection result in the cheek, it is possible to accurately determine a video in which a person's face has been synthesized.

[0132] In this embodiment, the video distribution device 10 may extract multiple regions representing different parts of a person and correct the score based on the emotion determination result in each region. It is not usually possible for different emotions to be determined in multiple regions representing different parts of a person. Therefore, according to this embodiment, it is possible to accurately determine the emotions in a video in which a person's face has been synthesized.

[0133] In this embodiment, the video distribution device 10 determines the emotion at each of multiple points in time. No change In such cases, the score value may be reduced. It is not usually possible for emotional fluctuations to not occur over a long period of time. Therefore, according to this embodiment, it is possible to accurately determine a video in which a person's face has been synthesized.

[0134] The video distribution system 1 in this embodiment displays a determination result indicating whether or not the video is a composite video, along with the video to be viewed. By referring to the determination result, the viewer can determine whether or not the video they are viewing is a composite video. Therefore, the video distribution system 1 in this embodiment can suppress the spread of composite videos generated by machine learning models.

[0135] [Second Embodiment] In the first embodiment, a configuration was described in which the video distribution device 10 determines whether or not the target video is a composite video and transmits the determination result along with the target video to the terminal device 20-2. In the second embodiment, a configuration is described in which the terminal device 20-2 determines whether or not the target video is a composite video and displays the determination result along with the target video.

[0136] Note that the terminal device 20-2 in this embodiment is an example of a determination device.

[0137] The following description will focus on the differences between the video distribution system 1 in this embodiment and the video distribution system 1 in the first embodiment.

[0138] <Functional Configuration> The functional configuration of the video distribution system 1 in this embodiment will be described with reference to Figure 16. Figure 16 is a block diagram showing an example of the functional configuration of the video distribution system 1 in this embodiment.

[0139] ≪Video Streaming Device≫ As shown in Figure 16, the video distribution device 10 in this embodiment comprises a video storage unit 100, a video receiving unit 101, a request receiving unit 102, and a video distribution unit 107. In other words, the video distribution device 10 in this embodiment differs from the video distribution device 10 in the first embodiment in that it does not include a video acquisition unit 103, a region extraction unit 104, a pulse wave detection unit 105, and a video determination unit 106.

[0140] In this embodiment, the video distribution unit 107 reads the target video from the video storage unit 100 based on the viewing request received by the request receiving unit 102. The video distribution unit 107 then transmits the read target video to the terminal device 20-2.

[0141] ≪Terminal Devices≫ As shown in Figure 16, the terminal device 20-2 in this embodiment includes a video acquisition unit 103, a region extraction unit 104, a pulse wave detection unit 105, a video determination unit 106, a video request unit 202, and a video display unit 203. In other words, the terminal device 20-2 in this embodiment differs from the terminal device 20-2 in the first embodiment in that it further includes a video acquisition unit 103, a region extraction unit 104, a pulse wave detection unit 105, and a video determination unit 106.

[0142] The region extraction unit 104, pulse wave detection unit 105, and video determination unit 106 provided in the terminal device 20-2 in this embodiment are the same as the region extraction unit 104, pulse wave detection unit 105, and video determination unit 106 provided in the video distribution device 10 in the embodiment.

[0143] In this embodiment, the video acquisition unit 103 receives the target video from the video distribution device 10.

[0144] In this embodiment, the video display unit 203 outputs the target video received by the video acquisition unit 103 and the determination result by the video determination unit 106 to the display device 506.

[0145] <Processing Procedure> The processing procedure of the video distribution method executed by the video distribution system 1 in this embodiment will be described with reference to Figure 17. Figure 17 is a sequence diagram showing an example of the processing procedure of the video distribution method in this embodiment.

[0146] Steps S31 to S33 are the same as steps S1 to S3 of the video distribution method in the first embodiment (see Figure 6).

[0147] In step S34, the request receiving unit 102 of the video distribution device 10 receives a video viewing request from the terminal device 20-2. Next, the request receiving unit 102 sends the identification information included in the video viewing request to the video distribution unit 107.

[0148] The video distribution unit 107 receives identification information from the request receiving unit 102. Next, the video distribution unit 107 identifies the target video stored in the video storage unit 100 based on the identification information. Subsequently, the video distribution unit 107 reads the target video from the video storage unit 100. Then, the video distribution unit 107 transmits the read target video to the terminal device 20-2.

[0149] In step S35, the video acquisition unit 103 of the terminal device 20-2 receives the target video from the video distribution device 10. The video acquisition unit 103 sends the received target video to the region extraction unit 104 and the video display unit 203.

[0150] In step S36, the region extraction unit 104 of the terminal device 20-2 receives the target video from the video acquisition unit 103. Next, the region extraction unit 104 extracts the measurement region included in the target video. Then, the region extraction unit 104 sends the measurement region information indicating the measurement region to the pulse wave detection unit 105.

[0151] In step S37, the pulse wave detection unit 105 of the terminal device 20-2 receives measurement area information from the area extraction unit 104. Next, the pulse wave detection unit 105 identifies the measurement areas in the target video based on the measurement area information. Subsequently, the pulse wave detection unit 105 extracts pulse wave signals in each identified measurement area. Then, the pulse wave detection unit 105 sends the pulse wave signals in each measurement area to the video determination unit 106.

[0152] In step S38, the video determination unit 106 of the terminal device 20-2 receives pulse wave signals from the pulse wave detection unit 105 in each measurement area. Next, the video determination unit 106 determines whether or not the target video is a composite video based on each pulse wave signal. Then, the video determination unit 106 sends the determination result regarding the target video to the video display unit 203.

[0153] In step S39, the video display unit 203 of the terminal device 20-2 receives the target video from the video acquisition unit 103. The video display unit 203 also receives the judgment result regarding the target video from the video judgment unit 106. Next, the video display unit 203 outputs the target video and the judgment result to the display device 506.

[0154] <Effects of the second embodiment> In this embodiment, the terminal device 20-2 detects pulse waves from a region indicating a predetermined part of a person included in the video, and determines whether the video is a composite video generated by a machine learning model based on the pulse wave detection result. Therefore, according to this embodiment, it is possible to determine whether a video in which a person is filmed is a composite video.

[0155] The determination process in this embodiment determines whether or not a video is a composite video based on pulse waves detected from the video, and therefore can be executed even on computers with limited computing resources, such as personal computers, smartphones, or tablet terminals. Accordingly, according to this embodiment, the determination process can be performed on the terminal device 20-2 that views the video.

[0156] [Differentiation] In each of the embodiments described above, a configuration was described in which it is determined whether or not the target video is a composite video after a request to view the target video has been made. The timing for determining whether or not the target video is a composite video is not limited to the above.

[0157] For example, it may be determined whether a target video is a composite video before a request to view the target video is made. For example, when a video is posted to the video distribution device 10, it may be determined whether the video is a composite video, and the determination result may be associated with the video and stored in the video storage unit 100. In this case, when a request to view the target video is made, the video distribution device 10 should distribute the determination result along with the target video.

[0158] In the second embodiment, a configuration was described in which the terminal device 20-2 determines whether or not a video distributed by a video distribution service is a composite video. The video to be determined is not limited to videos distributed by a video distribution service.

[0159] For example, video received by the terminal device 20-2 via email or messaging services may be included in the determination. Alternatively, video input to the terminal device 20-2 via portable storage media such as DVD-ROMs or USB memory may also be included in the determination.

[0160] Furthermore, for example, video received in real time by the terminal device 20-2 via a video call or conferencing system may be used as the target for judgment. When video received in real time is used as the target for judgment, instead of the average truth value of the entire target video, the truth value calculated at a predetermined time interval (for example, about 10 seconds) may be output as the judgment result.

[0161] [supplement] Each of the embodiments described above can be implemented by one or more processing circuits. Hereinafter, "processing circuit" as used herein includes processors programmed to execute each function by software, such as processors implemented by electronic circuits, as well as devices such as ASICs (Application Specific Integrated Circuits), DSPs (Digital Signal Processors), FPGAs (Field Programmable Gate Arrays), and conventional circuit modules designed to execute each of the functions described above.

[0162] While embodiments of the present disclosure have been described in detail above, the embodiments disclosed herein are illustrative and not restrictive in all respects. The embodiments can be modified and improved in various ways without departing from the scope and spirit of the appended claims. The features described in the above embodiments can be combined in any way that is not inconsistent with other configurations. [Explanation of Symbols]

[0163] 1. Video distribution system 10 Video distribution device 100 Video Memory Unit 101 Video Receiver 102 Request Receiving Unit 103 Video Acquisition Section 104 Area extraction part 105 Pulse wave detection unit 106 Video Judgment Unit 107 Video Streaming Department 20 Terminal devices 201 Video Transmission Section 202 Video Request Section 203 Video Display Section 401 Face Recognition Unit 402 Skin Judgment Department 403 Area determination section 601 Interval detection unit 602 Periodicity determination section 603 Chaos Analysis Department 604 Emotion Judgment Department 605 Truth calculation section 606 Authenticity storage unit 607 Result Output Section

Claims

1. A video acquisition unit configured to acquire video footage of a person, A region extraction unit configured to extract a region showing a predetermined part of the person included in the aforementioned video, A pulse wave detection unit configured to detect pulse waves from the aforementioned region, A video determination unit is configured to determine whether the video is a synthesized video generated by a machine learning model based on the pulse wave detection result, Equipped with, The region extraction unit is configured to extract multiple regions that represent different parts of the person, The video determination unit is configured to calculate a score indicating whether or not the video is a composite video based on the pulse wave detection results in each of the plurality of regions. Judgment device.

2. A determination device according to claim 1, The video determination unit is configured to determine that the video is a composite video when no pulse wave is detected in the region. Judgment device.

3. A determination device according to claim 2, The aforementioned area is the cheek of the person. Judgment device.

4. A determination device according to claim 1, The video analysis unit is configured to correct the score based on the detection result of the pulse wave in the person's cheek using the detection result of the pulse wave in the person's forehead or neck. Judgment device.

5. A determination device according to claim 4, The smaller the score, the more the video is a composite video. The video determination unit is configured to increase the score value when a pulse wave is detected from the region representing the forehead, and to decrease the score value when a pulse wave is not detected from the region representing the forehead. Judgment device.

6. A determination device according to claim 5, The video analysis unit is configured to reduce the score value when no pulse wave is detected from the region representing the neck. Judgment device.

7. A determination device according to claim 1, The system further includes an emotion determination unit configured to determine the emotions of the person based on the pulse wave, The video analysis unit is configured to correct the score based on the emotion analysis results in each of the multiple regions. Judgment device.

8. A determination device according to claim 7, The smaller the score, the more the video is a composite video. The video analysis unit is configured to reduce the score value when the result of determining the emotion on the person's cheek differs from the result of determining the emotion on the person's neck. Judgment device.

9. A determination device according to claim 8, The emotion determination unit is configured to determine the emotion at multiple points in time, The video analysis unit is configured to reduce the score value when the emotion analysis result at each of the multiple time points does not change. Judgment device.

10. A video distribution system in which terminal devices and video distribution devices can communicate via a network, The aforementioned video distribution device is A video acquisition unit configured to acquire video footage of a person from a storage unit in response to a request from the aforementioned terminal device, A region extraction unit configured to extract a region showing a predetermined part of the person included in the aforementioned video, A pulse wave detection unit configured to detect pulse waves from the aforementioned region, A video determination unit is configured to determine whether the video is a synthesized video generated by a machine learning model based on the pulse wave detection result, A video distribution unit configured to transmit the aforementioned video and the judgment result from the video judgment unit to the terminal device, Equipped with, The aforementioned terminal device is A video request unit configured to request the aforementioned video from the video distribution device, When the determination result indicates that the video is a composite video, a video display unit displays the video along with information indicating that the video is a composite video. Equipped with, The region extraction unit is configured to extract multiple regions that represent different parts of the person, The video determination unit is configured to calculate a score indicating whether or not the video is a composite video based on the pulse wave detection results in each of the plurality of regions. Video streaming system.

11. Computers The procedure for obtaining a video of a person, A procedure for extracting a region showing a specific part of the person contained in the aforementioned video, A procedure for detecting pulse waves from the aforementioned region, A procedure for determining whether the video is a synthesized video generated by a machine learning model based on the pulse wave detection result, Execute, The extraction procedure involves extracting multiple regions that represent different parts of the person, The procedure for making the determination involves calculating a score indicating whether or not the video is a composite video, based on the pulse wave detection results in each of the multiple regions. Judgment method.

12. On the computer, The procedure for obtaining a video of a person, A procedure for extracting a region showing a specific part of the person contained in the aforementioned video, A procedure for detecting pulse waves from the aforementioned region, A procedure for determining whether the video is a synthesized video generated by a machine learning model based on the pulse wave detection result, Make it run, The extraction procedure involves extracting multiple regions that represent different parts of the person, The procedure for making the determination involves calculating a score indicating whether or not the video is a composite video, based on the pulse wave detection results in each of the multiple regions. program.

Images