Sound transmission system, method, and program

Abstract

To provide a technique for ensuring discriminability between artificial voice and human voice in a system to transmit artificial voice generated by a mechanical learning model to a user of a vehicle.SOLUTION: The present disclosure provides a voice transmission system. The voice transmission system includes one or a plurality of processors. The one or the plurality of processors generate artificial voice resembled to human voice by a mechanical learning model, and then transmit the generated artificial voice to a user of a vehicle. At this time, the processors change a second parameter related to characteristics of the human voice in conjunction with a first parameter related to at least one of the movement of the vehicle and an operation amount by a driver of the vehicle.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] The present disclosure relates to a technique for transmitting voice to a user of a vehicle, and more particularly, to a technique for transmitting an artificial voice generated by a machine learning model. [Background technology]

[0002] Patent Document 1 discloses an in-vehicle device that reproduces a voice for a user in a vehicle based on speech voice information data received from a roadside device. The in-vehicle device includes a speech rate determination means and a voice signal generation means. The speech rate determination means determines a speech rate according to a situation when reproducing a voice from the received speech voice information data. The voice signal generation means generates a voice signal based on the speech voice information data so as to achieve the determined speech rate. Then, a voice based on the generated voice signal is output to a user from a speaker or the like. According to the speech rate determination means, the speech rate is set high when the speech voice information data is of high urgency in order to more reliably alert the user.

[0003] In this manner, a device for transmitting voice to a vehicle user is known. In addition to the above Patent Document 1, the following Patent Documents 2 and 3 can be cited as examples of documents that show the technical level at the time of filing of the present disclosure in the technical field. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2007-279975 A [Patent Document 2] JP 2001-033256 A [Patent Document 3] JP 2004-046400 A Summary of the Invention [Problem to be solved by the invention]

[0005] Consider applying a machine learning model to a device that transmits voice to a vehicle user, and outputting an artificial voice generated by the machine learning model to the user. By applying a machine learning model, it is possible to make the artificial voice closer to a live human voice, but it may be difficult for a user listening to the artificial voice to distinguish the artificial voice from a real human voice. However, from the viewpoint of transparency to a system to which a machine learning model is applied, it is desirable for the user to easily distinguish between an artificial voice generated by a machine learning model and a human voice.

[0006] One objective of the present disclosure is to provide a technology for ensuring ease of distinction between an artificial voice and a human voice in a system that transmits an artificial voice generated by a machine learning model to a vehicle user. [Means for solving the problem]

[0007] The present disclosure provides a voice transmission system. The voice transmission system of the present disclosure includes one or more processors. The one or more processors generate an artificial voice that resembles a human voice using a machine learning model. The generated artificial voice is then transmitted to a user of the vehicle. At this time, the processor changes a second parameter related to the nature of the artificial voice in conjunction with a first parameter related to at least one of the motion of the vehicle and the operation amount of the driver of the vehicle.

[0008] In the voice transmission system of the present disclosure, the first parameter may be an acceleration of the vehicle, and the second parameter may be a pitch of the artificial voice, and the processor may link the first parameter and the second parameter such that the pitch of the artificial voice increases as the acceleration increases.

[0009] The present disclosure provides a voice transmission method. The voice transmission method of the present disclosure includes the following steps. The first step is a step of generating an artificial voice that resembles a human voice using a machine learning model. The second step is a step of transmitting the artificial voice to a user of the vehicle. The third step is a step of changing a second parameter related to the properties of the artificial voice in conjunction with a first parameter related to at least one of the motion of the vehicle and the operation amount of the driver of the vehicle.

[0010] The present disclosure provides a voice transmission program executed by a computer, which causes a computer to execute the above-mentioned voice transmission method. Effect of the Invention

[0011] According to the technology of the present disclosure, the parameters related to the properties of the artificial voice are changed in conjunction with the parameters related to the vehicle motion or the amount of operation of the vehicle driver. In this way, voice changes in conjunction with the parameters related to the vehicle do not occur in voices emitted by real humans, so it is possible to prevent the user of the vehicle from mistaking the artificial voice emitted by the voice transmission system for a real human voice. In addition, there is no need to make the artificial voice sound mechanical in order to prevent the user from mistaking the artificial voice, so the sense of discomfort felt by the user from the artificial voice can be suppressed.

[0012] In this way, the technology disclosed herein makes it possible to make an artificial voice generated by a machine learning model easier to distinguish from an actual human voice while reducing the sense of discomfort felt by the user. [Brief description of the drawings]

[0013] [Figure 1] 1 is a block diagram showing an example of the configuration of a vehicle according to an embodiment of the present invention; [Diagram 2] 1 is a conceptual diagram showing an example of a scene in which voice is transmitted by the voice transmission system according to the present embodiment. FIG. [Diagram 3] FIG. 2 is a block diagram showing an example of processing by the audio transmission system. [Figure 4] 11 is a table showing examples of combinations of first and second parameters. [Diagram 5] FIG. 13 is a block diagram showing another example of processing by the audio transmission system. [Figure 6] FIG. 13 is a block diagram showing yet another example of processing by the audio transmission system. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0014] Embodiments of the present disclosure will be described with reference to the accompanying drawings.

[0015] 1. Overview of the voice transmission system The voice transmission system according to the present embodiment is a system that generates an artificial voice by a machine learning model and transmits the generated artificial voice to a vehicle user. The user to whom the artificial voice is transmitted may be the driver of the vehicle, or a passenger other than the driver. FIG. 1 is a diagram showing a configuration example of a vehicle 1 to which the voice transmission system 100 according to the present embodiment is applied. The vehicle 1 may be a vehicle owned by an individual, or may be a shared vehicle used for various mobility services. The vehicle 1 may also be a vehicle that is automatically driven by an automatic driving device.

[0016] The vehicle 1 includes a computer 10 that generates an artificial voice, a sensor group 20, and a speaker 30. The computer 10 is connected to the sensor group 20 and the speaker 30 via an in-vehicle network. The voice transmission system 100 includes at least the computer 10. However, the voice transmission system 100 may include the speaker 30 in addition to the computer 10. The voice transmission system 100 may further include the sensor group 20.

[0017] The computer 10 is, for example, an ECU (Electronic Control Unit) mounted on the vehicle 1, or a collection of multiple ECUs. Alternatively, some or all of the functions of the computer 10 may be located in an external server. In this case, the vehicle 1 and the external server are connected via a wireless communication network. In any case, the computer 10 includes one or more processors 11 (hereinafter simply referred to as processor 11) and one or more storage devices 12 (hereinafter simply referred to as storage devices 12).

[0018] The processor 11 executes various processes. Examples of the processor 11 include a CPU (Central Processing Unit), an ASIC (Application Specific Integrated Circuit), and an FPGA (Field-Programmable Gate Array). The processor 11 may be one type of processor, or may include multiple types of processors.

[0019] Examples of the storage device 12 include a hard disk drive (HDD), a solid state drive (SSD), a volatile memory, and a non-volatile memory. The storage device 12 may be one type of storage device, or may include multiple types of storage devices. The storage device 12 includes at least a program storage area 13 and a model data storage area 14. Each storage area may be realized by a single storage device, or may be realized by separate storage devices.

[0020] The program memory area 13 stores one or more programs. Each program is composed of a plurality of instructions. Various processes are realized by the voice transmission system 100 through cooperation between the processor 11 that executes the programs and the storage device 12. The programs stored in the program memory area 13 include a voice transmission program for causing the processor 11 to execute at least processes related to the transmission of artificial voice. The voice transmission program may be recorded on a computer-readable recording medium.

[0021] Model data is stored in the model data storage area 14. The model data stored in the model data storage area 14 is data of a machine learning model used to generate an artificial voice.

[0022] The storage device 12 may include a storage area other than the program storage area 13 and the model data storage area 14. In the storage area other than the program storage area 13 and the model data storage area 14, for example, map information is stored.

[0023] The sensor group 20 includes a vehicle state sensor 21. The vehicle state sensor 21 detects the state of the vehicle 1. Specifically, the vehicle state sensor 21 detects parameters such as the speed, acceleration, steering angle, roll angular velocity, pitch angular velocity, yaw angular velocity, and tilt angle of the vehicle 1. Examples of the vehicle state sensor 21 include a vehicle speed sensor, an acceleration sensor, a yaw rate sensor, a steering angle sensor, a roll angular velocity sensor, a pitch angular velocity sensor, a yaw angular velocity sensor, and a tilt angle sensor. The sensor group may also include a recognition sensor 22 and a position sensor 23. The recognition sensor 22 recognizes the situation around the vehicle 1. Examples of the recognition sensor 22 include a camera, LiDAR (Laser Imaging Detection and Ranging), and radar. Examples of the camera include a front camera that captures an image in front of the vehicle 1, a back camera that captures an image behind the vehicle 1, and a side camera that captures an image to the side of the vehicle 1. The position sensor 23 detects the position and orientation of the vehicle 1. An example of the position sensor 23 is a Global Navigation Satellite System (GNSS) sensor.

[0024] The computer 10 acquires information detected by these sensors through the in-vehicle network. For example, the computer 10 acquires the state of the vehicle 1 from a vehicle state sensor 21. The computer 10 may also acquire information detected by a recognition sensor 22 and a position sensor 23. Furthermore, the computer 10 may perform processing such as detection of objects present around the vehicle 1 and measurement of the relative position and relative speed of the detected object with respect to the vehicle 1, based on the information obtained from the recognition sensor 22.

[0025] The speaker 30 is a speaker that transmits the artificial voice generated by the computer 10 to the user of the vehicle 1. The speaker 30 may include multiple speakers. Data of the artificial voice generated by the computer 10 is sent to the speaker 30 through the in-vehicle network. Then, the speaker 30 reproduces the voice based on the received data, thereby transmitting the artificial voice to the user.

[0026] Above, we have described an example of the configuration of the voice transmission system 100. Next, we will explain the transmission of an artificial voice by the voice transmission system 100 by taking a specific example of a scene. Fig. 2 is a conceptual diagram showing an example of a scene in which an artificial voice is transmitted.

[0027] In the example of Fig. 2, the speaker 30 includes three speakers, a speaker 30a, a speaker 30b, and a speaker 30c. Among the three speakers, the speaker 30a is used to transmit artificial sound to the user. In this way, when the speaker 30 includes multiple speakers, only one speaker may emit sound at a time. Alternatively, sound may be emitted from multiple speakers at the same time.

[0028] In the example shown in FIG. 2, the vehicle 1 is an autonomous vehicle, and a voice message saying "Automatic driving will start" is transmitted to inform the user that the autonomous driving of the vehicle 1 will start. In this way, information can be sent to the user of the vehicle 1 by transmitting an artificial voice by the voice transmission system 100. Other examples of the contents of the transmitted artificial voice include messages such as "We will start moving," "We will change lanes," "We will arrive at the destination soon," "We are approaching a sharp curve," "There is an intersection ahead. Please be careful of pedestrians jumping out," "We are approaching a vehicle ahead. Please keep a sufficient distance between the vehicle," and "We have entered a road with a different speed limit. We will change the vehicle speed." However, the transmitted artificial voice is not limited to one that provides information, and may be one that calls some kind of attention, such as a voice that is emitted to call attention when drowsiness of the user is detected.

[0029] In order to convey these messages, the computer 10 uses a machine learning model to generate an artificial voice that resembles a human voice. Here, recent technological advances have made it possible to make the artificial voice generated using the machine learning model closer to a real human voice. By making the artificial voice closer to a real human voice, it is possible to reduce the sense of incongruity felt by the user receiving the voice, compared to when a mechanical voice is reproduced. On the other hand, it may be difficult for the user to distinguish whether the voice is an artificial voice or a real human voice when hearing it. For example, in the scene of FIG. 2, the driver of the vehicle 1 who hears the voice coming from the speaker 30a may mistakenly recognize the voice as being made by another person in the vehicle 1, for example, a passenger sitting in the passenger seat. Or, the driver may mistakenly recognize the voice as being transmitted via wireless communication from a person outside the vehicle 1.

[0030] From the viewpoint of transparency for a system to which a machine learning model is applied, it is desirable that it is easy for the user of the vehicle 1 to distinguish the artificial voice generated by the machine learning model from an actual human voice. However, in order to reduce the sense of incongruity felt by the user when hearing the artificial voice, the artificial voice itself needs to be close to a human voice, not a mechanical voice.

[0031] People can determine whether a voice they hear is a real human voice or an artificial voice by the naturalness of the changes in the voice characteristics. Voice characteristics, such as volume, pitch, and sound quality, change even in everyday conversations between real people, but these changes do not follow any rules. For example, factors that cause changes in the characteristics of a human voice in a conversation include the topic of conversation, the attitude of the conversation partner, and changes in the surrounding environment, but these do not give any regularity to the changes in the voice characteristics. Therefore, if there is some regularity in the changes in the voice that people hear, they will instinctively feel that the changes in the voice are unnatural and will unconsciously determine that it is not a real human voice, that is, an artificial voice.

[0032] Therefore, the voice transmission system 100 according to the present embodiment acquires a parameter related to the motion of the vehicle as a first parameter. Then, a second parameter related to the nature of the artificial voice is changed so as to be linked to the first parameter. By regularly changing the second parameter in linkage with the first parameter, the artificial voice transmitted by the voice transmission system 100 can be easily distinguished from an actual human voice.

[0033] 2. Processing example 3 is a block diagram showing an example of processing executed by the processor 11 as a result of execution of the voice transmission program. By executing the voice transmission program, the processor 11 executes processing 111, processing 112, processing 113, processing 114, and processing 115.

[0034] First, the processor 11 executes process 111. Process 111 detects a scene in which an artificial voice is to be transmitted. The scene in which an artificial voice is to be transmitted is, for example, a scene in which automatic driving of the vehicle 1 starts. In this case, the processor 11 determines the timing for automatic driving to start based on information received from the automatic driving device 200 that performs automatic driving of the vehicle 1. Then, when it is determined that it is the timing for automatic driving to start, the processor 11 detects the scene as a scene in which an artificial voice is to be transmitted.

[0035] The processor 11 may detect a scene in which an artificial voice is to be transmitted based on information received from a device other than the automatic driving device 200, for example, the recognition sensor 22. For example, the processor 11 calculates the distance between the vehicle 1 and the preceding vehicle based on information obtained by the recognition sensor 22 recognizing the preceding vehicle. Then, when the inter-vehicle distance between the vehicle 1 and the preceding vehicle is smaller than a predetermined distance, the processor 11 detects this as a scene in which an artificial voice is to be transmitted.

[0036] Alternatively, the processor 11 may detect a scene in which an artificial voice is to be transmitted based on map information stored in the storage device 12 and position information detected by the position sensor 23. For example, the processor 11 acquires the current position of the vehicle 1 from the position sensor 23. The processor 11 also acquires the destination of the vehicle 1 registered in the map information. Then, when it is determined that the vehicle 1 is approaching the destination based on this information, the processor 11 detects this as a scene in which an artificial voice is to be transmitted. As another example, when the vehicle 1 is approaching an intersection or a sharp curve, the processor 11 can similarly detect a scene in which an artificial voice is to be transmitted based on the map information and the position information of the vehicle 1.

[0037] The processor 11 may use a machine learning model to perform the process 111. That is, a scene in which an artificial voice is transmitted may be detected by the machine learning model.

[0038] Next, the processor 11 executes a process 112. The process 112 determines the content of the utterance by the artificial voice. The content of the artificial voice is determined according to the scene detected in the process 111. The content of the artificial voice corresponding to each scene is stored in advance in the storage device 12, for example. Alternatively, the processor 11 may execute the process 112 by using a machine learning model. That is, the content of the utterance by the artificial voice may be determined by the machine learning model.

[0039] The processor 11 executes processes 113 and 114 in parallel with processes 111 and 112. Process 113 acquires a parameter related to the motion of the vehicle 1 as a first parameter. Examples of the parameter related to the motion of the vehicle 1 include the speed, acceleration, steering angle, yaw rate, roll angular velocity, pitch angular velocity, yaw angular velocity, etc. of the vehicle 1. The processor 11 can acquire these parameters from the vehicle state sensor 21.

[0040] Next, the processor 11 executes process 114. Process 114 determines a second parameter. The second parameter is a parameter related to the property of the artificial voice generated by the voice transmission system 100. Examples of the second parameter include parameters representing the loudness (volume, sound pressure, amplitude), pitch (frequency), sound quality (waveform), etc. of the artificial voice. The second parameter is determined so as to be linked to the first parameter acquired in process 113.

[0041] Examples of combinations of the first and second parameters are shown in the table of Fig. 4. For example, the processor 11 links the acceleration of the vehicle 1 with the pitch of the artificial voice, and determines the pitch of the artificial voice so that the higher the acceleration of the vehicle 1, the higher the voice. Alternatively, for example, the processor 11 links the speed of the vehicle 1 with the volume of the artificial voice, and determines the volume of the artificial voice so that the higher the speed of the vehicle 1, the higher the voice. Alternatively, for example, the processor 11 links the yaw rate of the vehicle 1 with the sound quality of the artificial voice, and changes the sound quality of the artificial voice as the yaw rate of the vehicle 1 increases or decreases.

[0042] 4 are merely examples, and any parameters can be selected and combined as the first and second parameters. The parameters to be linked do not need to be combined one-to-one, and multiple second parameters may be linked to one first parameter. For example, both the pitch and volume of the artificial voice may be changed to be linked to the acceleration of the vehicle 1.

[0043] The combination of the first parameter and the second parameter and the correspondence indicating how the two parameters are linked are stored in advance in, for example, the storage device 12. Alternatively, the processor 11 may perform the process 114 by using a machine learning model. That is, the second parameter may be determined by the machine learning model so as to be linked with the first parameter.

[0044] Referring again to FIG. 3, after executing the process 112 and the process 114, the processor 11 executes the process 115. The process 115 generates an artificial voice based on the content determined in the process 112 and the parameters determined in the process 114. In the process 115, the processor 11 generates an artificial voice that resembles a human voice using a machine learning model. Note that a method of synthesizing an artificial voice that reads a set line using a machine learning model is well known, so it will not be described here. Data of the generated artificial voice is sent to the speaker 30, and the artificial voice is output from the speaker 30.

[0045] Through the above process, the voice is transmitted by the voice transmission system 100. As shown in the above process, according to the voice transmission system 100 of this embodiment, when transmitting voice, the second parameter related to the nature of the artificial voice is changed so as to be linked with the first parameter related to the motion of the vehicle 1. This makes it easier for the user to determine that the voice transmitted from the voice transmission system 100 is an artificial voice. In other words, the occurrence of a change in the nature of the voice that does not occur in an actual human voice makes it easier for the user to notice that the voice emitted by the voice transmission system 100 is an artificial voice, and prevents the user from misrecognizing that the voice is a human voice.

[0046] Furthermore, according to the voice transmission system 100 of the present embodiment, the second parameter is not simply changed, but is changed in conjunction with the first parameter, which makes it easier for the user to determine that the voice transmitted by the voice transmission system 100 is an artificial voice.

[0047] The motion of the vehicle 1 that determines the first parameter can be felt by the user riding in the vehicle 1. Therefore, the user can easily notice that the change in the second parameter is linked to the first parameter. Furthermore, a change in the nature of the voice that is linked to a parameter related to the behavior of the vehicle 1 does not usually occur in human voice. Therefore, it is possible for the user to more easily understand that the change in the second parameter is not merely coincidental but is caused by the voice transmission system 100.

[0048] One method for making the artificial voice generated by the system easier to distinguish from an actual human voice is to change the artificial voice to a mechanical voice, making it more distant from a human voice. However, for a user receiving the voice, a voice closer to a human voice is more likely to be pleasant to the ear than a mechanical voice. In the voice transmission system 100 according to the present embodiment, it is possible to improve the ease of distinguishing an artificial voice from another human voice while keeping the artificial voice closer to a human voice, thereby ensuring the transparency of the system and increasing the satisfaction of the user at the same time.

[0049] 3. Variations 3-1. First modified example (modified example regarding the first parameter) As the first parameter, a parameter other than the parameter related to the motion of the vehicle 1 may be used. For example, a parameter related to the operation amount of the driver of the vehicle 1 can be the first parameter. FIG. 5 is a block diagram showing an example of processing by the voice transmission program when a parameter related to the operation amount of the driver of the vehicle 1 is the first parameter. Here, the computer 10 is connected to the input device 40 in addition to the sensor group 20 and the speaker 30 via an in-vehicle network. The input device 40 is a device into which the driver of the vehicle 1 inputs the operation amount. Examples of the operation amount input by the driver include an accelerator operation amount, a brake operation amount, a steering operation amount, etc. Examples of the input device 40 include an accelerator pedal, a brake pedal, a steering wheel, etc.

[0050] Process 221 and process 222 are similar to process 111 and process 112 in Fig. 3. Process 223 and process 224 are executed in parallel with process 121 and process 122, and a first parameter is acquired by process 223, also similar to Fig. 3. However, the first parameter acquired by process 223 is a parameter related to the operation amount of the driver of vehicle 1. Processor 11 acquires the driver's operation amount from input device 40 as the first parameter, or calculates the first parameter based on the acquired operation amount.

[0051] The first parameter may be a parameter that represents the amount of operation of the driver itself. For example, a parameter that represents the magnitude of the accelerator amount input by the driver to the input device 40 may be the first parameter.

[0052] Alternatively, the difference between the ideal operation amount and the operation amount actually input by the driver may be the first parameter. The processor 11 can calculate the ideal operation amount based on information acquired from the sensor group 20. For example, the operation amount that keeps the speed of the vehicle 1 constant may be the ideal operation amount. In this case, the processor 11 can calculate the ideal operation amount based on the acceleration of the vehicle 1 acquired from the vehicle state sensor 21. Alternatively, for example, the operation amount that keeps the inter-vehicle distance from the preceding vehicle constant may be the ideal operation amount. The processor 11 calculates the relative distance between the vehicle 1 and the preceding vehicle from information obtained by the recognition sensor 22 recognizing the preceding vehicle. Then, the accelerator operation amount that keeps the relative distance from the preceding vehicle constant can be the ideal operation amount. Alternatively, the processor 11 may calculate, as the ideal operation amount, an operation amount that takes into account a change in gradient based on the inclination angle of the vehicle 1 acquired from the vehicle state sensor 21, so that the accelerator operation amount is large when the vehicle 1 travels on a road surface with an upward gradient, and the brake operation amount is large when the vehicle 1 travels on a road surface with a downward gradient.

[0053] Setting the parameter related to the driver's operation amount as the first parameter means that the parameter related to the operation input by the driver, who is the user of the vehicle 1, is set as the first parameter. In other words, the first parameter is set as a parameter related to an amount that the driver can grasp by himself. Therefore, it is possible to make it easier for the driver to notice that the second parameter related to the artificial voice is changed in conjunction with the first parameter. In this way, it is possible to make it easier for the user to determine that the voice transmitted by the voice transmission system 100 is an artificial voice using a machine learning model, and the transparency of the system can be improved.

[0054] Furthermore, when the difference between the ideal operation amount and the operation amount actually input by the driver is set as the first parameter, the following effect is expected. For the sake of explanation, consider a specific example of a situation in which the voice transmission system 100 transmits an artificial voice to alert the driver that the accelerator operation amount is too small or too large. For example, when the vehicle 1 approaches an uphill slope, there is a situation in which the accelerator operation amount becomes insufficient without the driver realizing, causing the vehicle 1 to decelerate. In such a situation, it is assumed that the driver is alerted by an artificial voice to step on the accelerator.

[0055] Here, the first parameter is the difference between the ideal accelerator operation amount and the actually input accelerator operation amount, and the second parameter is the pitch of the artificial voice. The pitch of the artificial voice changes according to the difference between the driver's accelerator operation amount and the ideal accelerator operation amount. The voice transmission system 100 not only notifies the driver that the accelerator operation amount is insufficient by the content of the artificial voice, but also becomes able to inform the driver of the amount of accelerator operation that the driver should depress by changing the pitch of the artificial voice. In other words, it also has the effect of making it easier for the driver to intuitively grasp the insufficient operation amount.

[0056] As yet another example of the first parameter, the distance between the vehicle 1 and a target present in front of the vehicle 1 may be used. An example of a target present in front of the vehicle 1 is a leading vehicle.

[0057] However, it is preferable that the first parameter is a parameter that the user can sense. By making the first parameter a parameter that the user can easily sense the change of, the user can easily notice that the first parameter and the second parameter are linked, which is more effective in preventing the user from mistaking the artificial voice for a human voice. Therefore, the most suitable first parameter is a parameter related to the motion of the vehicle 1 or the amount of operation by the driver of the vehicle 1.

[0058] 3-2. Second modification (modification regarding the second parameter) As the second parameter, a parameter other than the parameter related to the property of the artificial voice may be used. For example, the speech rate of the artificial voice may be used as the second parameter. In this case, for example, it is assumed that the first parameter is the speed of the vehicle 1, and the first parameter and the second parameter are linked so that the speech rate increases as the vehicle speed increases.

[0059] Alternatively, when the speaker 30 includes multiple speakers, the second parameter may be a parameter representing the position of the speaker that reproduces the artificial voice. For example, consider a case where multiple speakers 30 are installed on the left, right, and center of the vehicle interior as shown in FIG. 2. In this case, the first parameter is the steering angle of the vehicle 1, and the second parameter is the position of the speaker that reproduces the artificial voice, and the steering angle and the position of the speaker can be linked so that the sound source moves left and right as the steering angle changes left and right. Specifically, it is assumed that the artificial voice is emitted from the right speaker 30a when the steering angle increases to the right, from the left speaker 30c when the steering angle increases to the left, and from the center speaker 30b when the steering angle is close to 0. This makes it easier for the user to understand the change in the second parameter that is linked to the first parameter, and improves the ease of identifying whether the voice transmitted from the voice transmission system 100 is an artificial voice.

[0060] 3-3. The third variant (speech communication as dialogue) In the above embodiment, an embodiment has been described in which the artificial voice transmitted to the user by the voice transmission system 100 is a voice that is transmitted unilaterally for the purpose of notifying the user of information, etc. However, the artificial voice transmitted by the voice transmission system 100 may be a voice intended for dialogue with the user.

[0061] An example of processing executed by the voice transmission program when voice for the purpose of dialogue with a user is transmitted is shown in Fig. 6. In this modification, the voice transmission system 100 includes a microphone 50 for collecting voice emitted by a user. When the processor 11 executes process 311, the user's voice is collected by the microphone 50, and the collected voice is analyzed. The processor 11 may perform voice analysis using a machine learning model.

[0062] Next, the processor 11 executes process 312. In process 312, the content of a dialogue to be responded to the user's voice analyzed in process 311 is determined by the machine learning model.

[0063] Processes 313, 314, and 315 are similar to processes 113, 114, and 115 in Fig. 3. Process 315 generates an artificial voice based on the content of the dialogue determined in process 312, and transmits it to the user through the speaker 30.

[0064] When the artificial voice is intended for conversation, there is a high possibility that the voice will be continuously transmitted for a certain period of time by the voice transmission system 100. This makes it easier for the user to notice the change in the second parameter of the artificial voice, and makes it easier to distinguish that the voice transmitted from the voice transmission system 100 is an artificial voice.

[0065] 3-4. Fourth variant (voice communication system as an automated driving system) The voice transmission system 100 may be a part of an automatic driving system of the vehicle 1. In this case, an automatic driving device 200 of the vehicle 1 is configured by a part or all of the processor 11 and the storage device 12. Note that the automatic driving device 200 may be a device that performs automatic driving of the vehicle 1 by using a machine learning model.

[0066] 3-5. The fifth variation (upper and lower limits of the second parameter) The second parameter may be provided with an upper and lower limit. By providing the upper and lower limits, it is possible to prevent the second parameter from changing excessively even if the first parameter changes significantly. In particular, when the second parameter is the volume or pitch of the artificial voice, it is effective to provide the upper and lower limits. That is, when the second parameter is the volume of the artificial voice, by providing the upper and lower limits to the second parameter, it is possible to prevent the volume of the artificial voice from becoming excessively loud or quiet in conjunction with the first parameter, which would make it difficult for the user to hear. Similarly, when the second parameter is the pitch of the artificial voice, it is possible to prevent the pitch of the artificial voice from becoming excessively high or low, which would make it difficult for the user to hear.

[0067] The upper and lower limits of the second parameter are set in advance in consideration of the ease of hearing for the user, and are stored in the storage device 12. When the second parameter exceeds the upper limit or falls below the lower limit, the second parameter will not change even if the first parameter changes more significantly than that. In this case, the nature of the artificial voice has changed to an unnatural level as a human voice, so it is considered that there is little risk that the user will mistakenly recognize the voice produced by the voice transmission system 100 as a human voice.

[0068] 4. Application Examples 4-1. First example Finally, a specific example of a scene to which the voice transmission system 100 according to the present embodiment is applied will be described. In the first example, a scene in which the vehicle 1 is traveling on a road that is inclined sideways is considered. In addition, the first parameter is the roll angle of the vehicle 1, and the second parameter is the sound quality of the artificial voice. When the vehicle speed of the vehicle 1 is too fast and the roll angle becomes large, and there is a risk of the vehicle 1 rolling over, it is assumed that the voice transmission system 100 transmits a voice to alert the driver. At this time, the driver can easily recognize that the voice transmitted from the voice transmission system 100 is an artificial voice due to the change in sound quality of the artificial voice linked to the roll angle of the vehicle 1. At the same time, the sound quality of the artificial voice changes in conjunction with the magnitude of the roll angle of the vehicle 1, so that the driver can feel the magnitude of the roll angle from the change in sound quality, and it becomes easier to recognize how much operation is required to return the vehicle to the correct posture.

[0069] 4-2. Second example As a second example, consider a case where the vehicle 1 is a vehicle towing a trailer, and the recognition sensor 22 includes a backup camera that captures an image of the rear of the vehicle 1. In this case, the first parameter may be a parameter related to the swing of the trailer towed by the vehicle 1, and the second parameter may be the volume of the artificial voice. As the parameter related to the swing of the trailer, for example, the speed, volume, period, etc. of the swing are set as the first parameter. Note that the processor 11 can acquire these parameters related to the swing of the trailer by image analysis of the image captured by the backup camera.

[0070] When the trailer towed by the vehicle 1 starts to swing left and right, it is assumed that the voice transmission system 100 transmits an artificial voice to the user informing the user that the trailer is swinging. At this time, for example, the volume of the artificial voice is linked to the magnitude of the swing so that the volume of the artificial voice becomes larger as the trailer swings larger. It becomes easier for the user to determine that the voice transmitted from the voice transmission system 100 is an artificial voice due to the change in the second parameter linked to the first parameter. At the same time, the volume of the artificial voice makes it easier to recognize the magnitude of the swing. In other words, the larger the swing becomes and the greater the importance of dealing with it becomes, the louder the volume of the artificial voice becomes, so that it is also possible to obtain the effect that the user can more easily grasp the importance of the situation notified by the artificial voice. [Explanation of symbols]

[0071] 1 vehicle, 10 computer, 11 processor, 12 storage device, 13 program storage area, 14 model data storage area, 20 sensor group, 21 vehicle state sensor, 22 recognition sensor, 23 position sensor, 30 speaker, 40 input device, 50 microphone, 100 voice transmission system, 200 automatic driving device

Claims

1. 1. An audio transmission system, comprising: One or more processors; The one or more processors: An artificial voice that resembles a human voice is generated using a machine learning model. transmitting the artificial voice to a user of the vehicle; A second parameter relating to the nature of the artificial voice is changed in conjunction with a first parameter relating to at least one of the motion of the vehicle and the operation amount of the driver of the vehicle.

1. A voice transmission system comprising:

2. 2. The audio transmission system of claim 1, The first parameter is any one of a speed, an acceleration, and a steering angle of the vehicle.

1. A voice transmission system comprising:

3. 3. The audio transmission system according to claim 1 or 2, The second parameter includes at least one of the volume, pitch, and sound quality of the artificial voice.

1. A voice transmission system comprising:

4. 4. The audio transmission system according to claim 3, the first parameter is the acceleration of the vehicle; the second parameter includes a pitch of the artificial voice; The machine learning model increases the pitch of the artificial voice as the acceleration increases.

1. A voice transmission system comprising:

5. 4. The audio transmission system according to claim 3, The second parameter has an upper limit and a lower limit.

1. A voice transmission system comprising:

6. 2. The audio transmission system of claim 1, The first parameter is This is the difference between the ideal operation amount and the operation amount actually input by the driver.

1. A voice transmission system comprising:

7. A method for transmitting audio, comprising: generating an artificial voice that resembles a human voice using a machine learning model; transmitting the artificial voice to a user of the vehicle; changing a parameter related to the characteristics of the artificial voice in conjunction with a parameter related to at least one of the motion of the vehicle and the operation amount of the driver of the vehicle; Includes 13. A method for transmitting voice, comprising:

8. A voice transmission program executed by a computer, generating an artificial voice that resembles a human voice using a machine learning model; transmitting the artificial voice to a user of the vehicle; changing a parameter related to the characteristics of the artificial voice in conjunction with a parameter related to at least one of the motion of the vehicle and the operation amount of the driver of the vehicle; Have your computer run A voice transmission program comprising:

Images