JP7485038B2 - Speech generation device, speech generation method, and speech generation program - Google Patents

Description

This embodiment relates to a voice generation device, a voice generation method, and a voice generation program.

It is difficult for a person to spend the day without feeling sleepy at all. This is because the human brain functions in a rhythm of fluctuations in alertness with short periods known as ultradian rhythms. Non-Patent Documents 1 and 2 explain that sleep is a state that is the polar opposite of alertness. The alertness level is an index that indicates the degree of alertness that goes from sleep to excitement. Non-Patent Document 2 defines "drowsiness" as a state in which the alertness level is lower than a moderate alertness level. When working in a remote environment such as working from home or attending remote classes, even if one feels sleepy, it is necessary to increase the alertness level as quickly as possible.

Masaharu Yoshio, “Level of Arousal”, Journal of Physical Therapy, Vol. 37, No. 12 Hiroki Kitajima et al., "Study on prediction method of drowsiness while driving: 1st report, Evaluation method of drowsiness facial expression and effective index for predicting drowsiness fluctuation", Transactions of the Japan Society of Mechanical Engineers, Series C, Vol. 63, No. 613, 1997, pp. 3059-3066 Tatsuya Ishida et al., “Proposal of a Cocktail Party Effect Support System by Speaker Identification Using Temperature Sensors and Its Feasibility Verification Experiments”, Proceedings of the 33rd Fuzzy Systems Symposium, p.373-376 Takuya Kume et al., "Development of a method to detect and eliminate distracted driving using in-vehicle devices," Transactions of the Society of Automotive Engineers of Japan, Vol. 45, No. 3, pp. 567-572

The embodiments provide a voice generation device, a voice generation method, and a voice generation program for waking up a user in a short period of time.

The voice generation device of the embodiment includes an acquisition unit that acquires a first level of alertness that represents the user's degree of alertness from sleep to excitement, a judgment unit that determines whether the user is in an alert state based on the first level of alertness, a selection unit that selects a voice that encourages the user to wake up from a plurality of voice candidates when the user is not in an alert state, and a generation unit that generates a voice call to present to the user based on the selected voice , wherein each of the plurality of voice candidates is associated with a familiarity level that represents the degree to which the user is familiar with the corresponding voice candidate, and the selection unit selects the voice that encourages the user to wake up in accordance with the familiarity level .

According to an embodiment, a voice generation device, a voice generation method, and a voice generation program are provided to wake up a user in a short period of time.

FIG. 1 is a diagram illustrating an example of a hardware configuration of a sound generating device according to an embodiment. FIG. 2 is a diagram illustrating an example of the configuration of the familiarity DB. FIG. 3 is a diagram illustrating an example of the configuration of the user log DB. FIG. 4 is a diagram showing an example of the configuration of the call-out DB. FIG. 5 is a diagram illustrating an example of the configuration of the active behavior DB. FIG. 6 is a functional block diagram of the voice generating device. FIG. 7 is a flowchart showing a voice presentation process performed by the voice generating device. FIG. 8 is a flowchart of the active behavior presentation process. FIG. 9 is a graph showing a selection function in which the awakening level is x and the required time is y.

Hereinafter, an embodiment will be described with reference to the drawings. FIG. 1 is a diagram showing a hardware configuration of an example of a voice generation device according to an embodiment. The voice generation device 1 according to the embodiment emits a voice prompt to wake up a user when the user is not in an alert state, such as when the user is feeling drowsy.

In the embodiment, whether or not the user is in an awakened state is determined based on the "awakening level". The awakening level in the embodiment is an index indicating the degree of awakening corresponding to the awakening level. The awakening level corresponds to the activity level of the cerebrum and represents the degree of awakening from sleep to excitement. The awakening level is measured from eye movement, blinking activity, electrodermal activity, reaction time to a stimulus, and the like. The awakening level in the embodiment is calculated from any one or a combination of eye movement, blinking activity, electrodermal activity, and reaction time to a stimulus for measuring these awakening levels. The awakening level is, for example, a value that increases as the state moves from a sleep state to an excited state. The awakening level may be a continuous numerical value or may be a discrete value such as Level 1, Level 2, .... In addition, when the awakening level is calculated by combining the values of eye movement, blinking activity, electrodermal activity, and reaction time to a stimulus, the method of combining them is not particularly limited. For example, the combination method may be a simple summation of these values, a weighted addition, or the like.

The voice generating device 1 has a processor 2, a ROM 3, a RAM 4, a storage 5, a microphone (mic) 6, a speaker 7, a camera 8, an input device 9, a display 10, and a communication module 11. The voice generating device 1 is, for example, a personal computer (PC), a smartphone, a tablet terminal, or any other terminal. The voice generating device 1 may be mounted on any device used by a user. The voice generating device 1 does not need to have all of the configurations shown in FIG. 1. For example, the microphone 6, the speaker 7, the camera 8, and the display 10 may be separate devices from the voice generating device 1.

Processor 2 is a control circuit that controls the overall operation of the voice generating device 1, such as a CPU. Processor 2 does not have to be a CPU, but may be an ASIC, FPGA, GPU, etc. Processor 2 does not have to be composed of a single CPU, but may be composed of multiple CPUs, etc.

ROM 3 is a non-volatile memory such as a flash memory. For example, a startup program for the voice generating device 1 is stored in ROM 3. RAM 4 is a volatile memory such as an SDRAM. RAM 4 can be used as a working memory for various processes in the voice generating device 1.

Storage 5 is a storage such as a flash memory, a hard disk drive (HDD), or a solid state drive (SSD). Various programs used by the voice generating device 1 are stored in storage 5. Storage 5 may also store a familiarity database (DB), a user log database (DB) 52, a voice synthesis parameter database (DB) 53, a call-out sentence database (DB) 54, and an active behavior database (DB) 55. These databases will be described in detail later.

The microphone 6 is a device that converts input voice into a voice signal, which is an electrical signal. The voice signal obtained by the microphone 6 can be stored, for example, in the RAM 4 or the storage 5. For example, voice synthesis parameters for synthesizing a voice call can be obtained from the voice input via the microphone 6.

Speaker 7 is a device that outputs sound based on the input audio signal.

The camera 8 captures an image of the user and acquires an image of the user. The image of the user acquired by the camera 8 may be stored, for example, in the RAM 4 or the storage 5. The image of the user is used, for example, to acquire the user's level of alertness or to acquire the user's response to a voice call.

The input device 9 is a mechanical input device such as a button, switch, keyboard, or mouse, or a software input device using a touch sensor. The input device 9 accepts various inputs from the user. The input device 9 then outputs a signal to the processor 2 in response to the user's input.

The display 10 is, for example, a liquid crystal display or an organic electroluminescence display. The display 10 displays various images.

The communication module 11 is a device for the voice generating device 1 to carry out communication. The communication module 11 communicates, for example, with a server provided outside the voice generating device 1. The method of communication by the communication module 11 is not particularly limited. The communication module 11 may carry out communication wirelessly or wired.

Next, we will explain the familiarity database (DB) 51, user log database (DB) 52, voice synthesis parameter database (DB) 53, call-out sentence database (DB) 54, and active behavior database (DB) 55.

Figure 2 is a diagram showing an example of the configuration of the familiarity DB51. The familiarity DB51 is a database that records the "familiarity" of the user. The familiarity DB51 records, for example, a user ID, a voice label, a familiarity target, a familiarity, the number of responses, the number of presentations, and the average arousal change value in association with each other.

"User ID" is an ID assigned to each user of the voice generating device 1. The user ID may be associated with user attribute information such as a user name.

A "voice label" is a label that is uniquely assigned to each candidate call voice. Any label may be used as the voice label. For example, the familiar name of the target may be used as the voice label.

A "familiar target" is a person with whom the user regularly converses or a target that emits a voice that the user frequently hears. A familiar target does not necessarily have to be a person.

"Familiarity" is the degree of familiarity of the user with the voice of the corresponding familiar target. Familiarity can be calculated from the frequency of communication with the familiar target via SNS etc., the frequency of daily conversation with the familiar target, the frequency of hearing the familiar target on a daily basis, etc. For example, the higher the frequency of communication with the familiar target via SNS etc., the frequency of daily conversation with the familiar target, and the frequency of hearing the familiar target on a daily basis, the higher the familiarity value. Here, familiarity may be obtained by self-reporting by the user.

The "number of responses" is the number of times the user responded to the call voice generated based on the corresponding voice label. The number of presentations is the number of times the call voice generated based on the corresponding voice label was presented to the user. The response probability can be calculated by dividing the number of responses by the number of presentations. The response probability is the probability that the user will respond to the call voice generated based on the corresponding voice label.

"Average change in alertness" is the average change in the user's alertness in response to the call voice generated based on the corresponding voice label. The change in alertness will be explained later.

Figure 3 is a diagram showing an example of the configuration of user log DB52. User log DB52 is a database that records logs related to the user's use of the voice generating device 1. User log DB52 records, for example, the date and time the log was generated, a user ID, a voice label, a familiarity target, whether or not there was a reaction, a level of alertness, a new level of alertness, and an amount of change in alertness, all in association with each other. The user ID, voice label, and familiarity target are the same as those in familiarity DB51.

The "log generation date and time" is the date and time when the user used the voice generating device 1. The log generation date and time is recorded, for example, each time a voice call is presented to the user.

"Response or no response" is information about whether or not the user responded after the voice prompt was presented to the user. If the user responded, "Yes" is recorded. If the user did not respond, "No" is recorded.

"Level of alertness" is the level of alertness obtained before the voice generating device 1 presents the voice call.

"New alertness" is the new alertness obtained after the user reacts. New alertness is not recorded when there is no user reaction.

"Change in alertness" is an amount that represents the change in alertness before and after the user's reaction. For example, the change in alertness can be obtained from the difference between the new alertness and the alertness. The change in alertness may also be the ratio between the new alertness and the alertness, etc. The change in alertness is not recorded when there is no reaction from the user.

The voice synthesis parameter DB53 is a database that records voice synthesis parameters. The voice synthesis parameters are data used to synthesize a voice that is familiar to the user. For example, the voice synthesis parameters may be feature data extracted from voice data previously collected via the microphone 6. Alternatively, voice synthesis parameters acquired or defined by another system may be recorded in advance. Here, the voice synthesis parameters are associated with voice labels.

Figure 4 is a diagram showing an example of the configuration of the call DB 54. The call DB 54 is a database that records template data of various call statements for waking up the user. There are no particular limitations on the call statements. However, it is desirable for the call statements to include a call using the user's name. This is to enhance the cocktail party effect, which will be explained later.

Figure 5 is a diagram showing an example of the configuration of active behavior DB 55. Active behavior DB 55 is a database that records information on active behaviors that can be presented to a user in list format. For example, active behavior DB 55 records an ID, an action content, a required time, and constraints in association with each other.

"ID" is an ID assigned to each active behavior. Here, an active behavior is an action that the user is made to actively perform in order to encourage the user to wake up.

"Behavior content" is the specific content of the active behavior. The behavior content may be recorded, for example, as text data.

"Time required" is the time required to perform the corresponding active behavior. Here, in Figure 5, the time required is recorded in seconds. The time required does not necessarily have to be recorded in seconds.

A "constraint" is a condition necessary to perform the corresponding active action. For example, "Space Required" in Figure 5 means that some space is required to perform the corresponding active action. The constraint may be recorded as text data, for example, or may be selected from a number of constraints that are registered in advance.

Here, the familiarity DB 51, the user log DB 52, the voice synthesis parameter DB 53, the call out sentence DB 54, and the active behavior DB 55 do not necessarily have to be stored in the storage 5. For example, the familiarity DB 51, the user log DB 52, the voice synthesis parameter DB 53, the call out sentence DB 54, and the active behavior DB 55 may be stored in a server separate from the voice generation device 1. In this case, the voice generation device 1 accesses the server using the communication module 11 and acquires the necessary information.

Figure 6 is a functional block diagram of the voice generation device 1. As shown in Figure 6, the voice generation device 1 has an acquisition unit 21, a determination unit 22, a selection unit 23, a generation unit 24, and a presentation unit 25. The operations of the acquisition unit 21, the determination unit 22, the selection unit 23, the generation unit 24, and the presentation unit 25 are realized, for example, by the processor 2 executing a program stored in the storage 5. The determination unit 22, the selection unit 23, the generation unit 24, and the presentation unit 25 may be realized by hardware separate from the processor 2.

The acquisition unit 21 acquires the user's level of alertness. The acquisition unit 21 also acquires the user's reaction to the voice call. As described above, the level of alertness is calculated from any one of eye movement, blink activity, skin electrodermal activity, and reaction time to a stimulus, or a combination thereof. Here, the eye movement, blink activity, and reaction time to a stimulus for calculating the level of alertness can be measured, for example, from an image of the user acquired by the camera 8. The reaction time to a stimulus may also be measured from an audio signal acquired by the microphone 6. The skin electrodermal activity can be measured, for example, by a sensor attached to the user's arm. The user's reaction can also be acquired by measuring, for example, from an image acquired by the camera 8, whether the user looked in the direction of the sound after the voice call is presented. The acquisition unit 21 may be configured to acquire the level of alertness or the user's reaction calculated outside the voice generating device 1 through communication.

The determination unit 22 determines whether or not the user is awake based on the alertness level acquired by the acquisition unit 21. When the determination unit 22 determines that the user is awake, it transmits a request to select a voice label to the receiving unit 231 of the selection unit 23. Here, the determination unit 22 makes the determination by comparing the alertness level with a predetermined threshold. The threshold is an alertness threshold for determining whether or not the user is awake, and is stored in the storage 5, for example. The determination unit 22 also determines whether or not the user has responded based on the information on the user's response acquired by the acquisition unit 21.

The selection unit 23 selects a voice label of a candidate voice for encouraging the user to wake up when it is determined that the user is not in an awake state. The selection unit 23 has a receiving unit 231, a voice label selecting unit 232, and a transmitting unit 233.

The receiving unit 231 receives a request to select an audio label from the judgment unit 22.

The voice label selection unit 232 selects a voice label by referring to the familiarity DB 51. People have the characteristic that when they hear a familiar voice or words, they suddenly feel that they can hear it. This characteristic is called the cocktail party effect. For example, many users have experienced being startled when they hear a familiar voice or when their name is called. This is also due to the cocktail party effect. In the embodiment, the cocktail party effect is utilized to call the user in a familiar voice, thereby promoting an increase in the user's level of alertness. The voice label selection unit 232 selects a voice that is easier for the user to perceive, i.e., a more familiar voice, based on the familiarity.

The transmission unit 233 transmits information about the voice label selected by the voice label selection unit 232 to the generation unit 24.

The generation unit 24 generates a voice call to wake up the user based on the voice label received from the transmission unit 233. The generation unit 24 acquires voice synthesis parameters corresponding to the voice label received from the transmission unit 233 from the voice synthesis parameter DB 53. The generation unit 24 then generates a voice call based on the data of the call sentence recorded in the call sentence DB 54 and the voice synthesis parameters.

The presentation unit 25 presents to the user the voice prompt generated by the generation unit 24. For example, the presentation unit 25 plays the voice prompt generated by the generation unit 24 using the speaker 7.

Next, the operation of the voice generating device 1 will be described. Figure 7 is a flowchart showing the voice presentation process by the voice generating device 1. The process in Figure 7 may be performed periodically.

In step S1, the acquisition unit 21 acquires the user's level of alertness. The acquisition unit 21 outputs the acquired level of alertness to the determination unit 22. The acquisition unit 21 also retains the acquired level of alertness until the timing of acquiring a response from the user after the presentation of the voice call.

In step S2, the determination unit 22 determines whether the level of alertness acquired by the acquisition unit 21 is equal to or lower than a threshold value. When it is determined in step S2 that the level of alertness exceeds the threshold value, i.e., the user is in an alert state, the processing in FIG. 7 ends. When it is determined in step S2 that the level of alertness is equal to or lower than the threshold value, i.e., the user is not in an alert state such as feeling drowsy, the processing proceeds to step S3.

In step S3, the determination unit 22 transmits a voice label selection request to the selection unit 23. When the voice label selection request is received by the receiving unit 231, the voice label selection unit 232 refers to the user log DB 52 and determines whether the number of times the user has used the voice generation device 1 is 0. The number of uses can be calculated, for example, from the number of log occurrence dates and times recorded in the user log DB 52. When it is determined in step S3 that the number of times the user has used the voice generation device 1 is 0, the process proceeds to step S4. When it is determined in step S3 that the number of times the user has used the voice generation device 1 is not 0, the process proceeds to step S6.

In step S4, the voice label selection unit 232 acquires the familiarity C _i corresponding to all candidate voice labels from the familiarity DB 51. Here, C _i is the familiarity of the voice label Li (i=1, 2, . . . ).

In step S5, the voice label selection unit 232 selects a voice label corresponding to the maximum familiarity among the obtained familiarities. That is, the voice label selection unit 232 selects a voice label according to the following conditional expression (1). After that, the process proceeds to step S8.
maxC _i Equation (1)

In step S6, the voice label selection unit 232 obtains the familiarity C _i , response probability R _i , and average awakening change A _i corresponding to all candidate voice labels from the familiarity DB 51. Here, C _i is the familiarity of voice label Li. R _i is the response probability of voice label Li. A _i is the average awakening change value of voice label Li. When the number of uses is not 0, the response probability and the average awakening change value are obtained. Therefore, when the number of uses is not 0, these response probabilities and the average awakening change value are also used to select a voice label.

In step S7, the voice label selection unit 232 selects a voice label corresponding to the combination of familiarity, response probability, and average arousal change value that maximizes the weighted sum of the obtained familiarity, response probability, and average arousal change value. That is, the voice label selection unit 232 selects a voice label according to the following conditional formula (2). After that, the process proceeds to step S8.
max( _{wcCi + wrRi + waAi )} Equation (2)
Here, _wc , _wr , and _wa in formula (2) are weights of C _i , R _i , and A _i , respectively. These _wc , _wr , and _wa may be predefined. Also, _wc , _wr , and _wa may be changed as appropriate. For example, while the number of times of use is small, _wc , _wr , and _wa are set to the same value, and thereafter, if it is better to emphasize any one of familiarity, response probability, and average value of change in arousal level, the weight may be changed accordingly.

In step S8, the transmission unit 233 transmits information indicating the voice label selected by the voice label selection unit 232 to the generation unit 24. The generation unit 24 acquires voice synthesis parameters corresponding to the received voice label from the voice synthesis parameter DB 53. The generation unit 24 then generates a call voice based on the data of a call sentence randomly selected from the call sentence DB 54 and the voice synthesis parameters. The call voice can be generated by a voice synthesis process using the voice synthesis parameters. The process then proceeds to step S9.

In step S9, the presentation unit 25 presents the call voice generated by the generation unit 24 to the user using the speaker 7.

In step S10, the acquisition unit 21 acquires the user's reaction. Then, the acquisition unit 21 outputs information on the user's reaction to the determination unit 22.

In step S11, the judgment unit 22 judges whether or not there has been a reaction from the user. If it is determined in step S11 that there has been no reaction from the user, the process proceeds to step S12. If it is determined in step S11 that there has been a reaction from the user, the process proceeds to step S13.

In step S12, the determination unit 22 transmits information to the selection unit 23 that the user has not responded. When the receiving unit 231 receives the information that the user has not responded, the voice label selection unit 232 selects the next candidate voice label. When the voice label selection unit 232 selects a voice label based on the formula (1), the voice label selection unit 232 selects a voice label corresponding to the next largest familiarity C _i . When the voice label selection unit 232 selects a voice label based on the formula (2), the voice label selection unit 232 selects a voice label corresponding to the next largest weighted sum value. In addition, the determination unit 22 transmits information to the acquisition unit 21 that the user has not responded. In response to this, the acquisition unit 21 registers the information on the no-response in the user log DB 52 in association with the log generation date and time, the voice label, the familiarity target, and the familiarity. After that, the process returns to step S8. In this case, the call voice is presented again. Note that, when all the candidate voice labels have been selected in step S12, the process of FIG. 7 may be terminated.

In step S13, the determination unit 22 requests the acquisition unit 21 to acquire the new awakening level. In response to this, the acquisition unit 21 acquires the new awakening level. The acquisition of the new awakening level may be performed in the same manner as the acquisition of the awakening level.

In step S14, the acquisition unit 21 calculates the amount of change in alertness diff(new alertness, alertness). As described above, the amount of change in alertness can be calculated, for example, from the difference between the new alertness and the alertness.

In step S15, the acquisition unit 21 acquires the average awakening change value from the familiarity DB 51. The acquisition unit 21 then updates the average awakening change value using the newly calculated amount of awakening change and the previously acquired average awakening change value. The acquisition unit 21 also registers in the user log DB 52 information on the presence of a response, the awakening level, the new awakening level, and the amount of awakening change in association with the log generation date and time, the voice label, the familiarity target, and the familiarity level.

In step S16, the judgment unit 22 acquires the new awakening level from the acquisition unit 21. The judgment unit 22 then judges whether or not the new awakening level is equal to or lower than a threshold value. The threshold value in step S16 may be the same as or different from the threshold value in step S2. When it is judged in step S16 that the new awakening level is not equal to or lower than the threshold value, the processing in FIG. 7 ends. When it is judged in step S16 that the new awakening level is equal to or lower than the threshold value, the processing proceeds to step S17. Note that the judgment in step S16 and the processing in step S17 may be performed in parallel with the processing in step S15.

In step S17, the determination unit 22 requests the generation unit 24 to perform active behavior presentation processing. In response to this, the generation unit 24 performs the active behavior presentation processing. After the active behavior presentation processing, the processing in FIG. 7 ends.

Figure 8 is a flowchart of the active behavior presentation process. In step S21, the generation unit 24 refers to the active behavior DB 55 to obtain a list of active behaviors available to the user. Note that it is not necessary to obtain a list of active behaviors for which the current user state does not match the constraints. For example, when the space required for an active behavior cannot be secured, information on active behaviors that require space as a constraint does not need to be obtained.

In step S22, the generation unit 24 calculates a selection function. The selection function is a function for selecting an active behavior. Below, an example of how to calculate the selection function is described.

The generation unit 24 acquires the required time max, the required time _{min, the alertness max, and the alertness min from the required time included in the list of acquired active behaviors and the alertness recorded in the user log DB 52. Here, the required time max and the required time min are the} maximum and minimum values of the required time of the acquired active behaviors, respectively. Also, the alertness _max and the alertness _min are the maximum and minimum values of the acquired alertness, respectively. Here, the alertness acquired may be only the alertness of the corresponding user. Alternatively, the alertness of all users may be acquired as the alertness of humans in general.

After acquiring the required time _max , the required time _min , the awakening level _max , and the awakening level _min , the generation unit 24 calculates a linear function y=f(x) that passes through the two points (awakening level _max , required time _min ) and (awakening level _min , required time _max ) as a selection function. Fig. 9 is a graph showing the selection function with the awakening level as x and the required time as y.

Now, we return to the explanation of FIG. 8. In step S23, the generation unit 24 extracts an active behavior that matches the new awakening level based on the selection function. For example, the generation unit 24 obtains the time required for the current user to awaken using the current new awakening level as an input to the selection function. Then, the generation unit 24 extracts an active behavior that is closest to the time required for the current user to awaken from the acquired list of active behaviors.

In step S25, the generation unit 24 generates a voice call suggesting active behavior. The generation of the voice call may be performed by a voice synthesis process using voice synthesis parameters. The voice synthesis parameters used at this time may be the voice synthesis parameters used to synthesize the voice call in the immediately preceding step S8. The voice synthesis parameters used in the immediately preceding voice synthesis are the voice synthesis parameters used to generate the voice call to which the user responded. Therefore, by using the voice synthesis parameters used to synthesize the voice call in the immediately preceding step S8, the user's response to the voice call suggesting active behavior is also expected.

In step S26, the presentation unit 25 presents the voice call generated by the generation unit 24 to the user using the speaker 7. After that, the process of Fig. 8 returns to the process of Fig. 7. When presenting the voice call suggesting an active behavior to the user, for example, an image representing the active behavior may be displayed on the display 10.

As described above, according to the embodiment, when it is determined that the user is not awake, the user is called out to using a voice familiar to the user. Therefore, even if the user is in a drowsy state, it is expected that the user will hear the voice call due to the cocktail party effect. Therefore, it is expected that the user's level of wakefulness will improve in a short period of time.

According to the embodiment, when the user uses the voice generating device for the second or subsequent time, a voice that is familiar to the user is selected in consideration of the user's reaction and the amount of change in arousal level in addition to the familiarity. Therefore, when the voice generating device is used for the second or subsequent time, an improvement in arousal level is expected in a shorter time than the first time.

Furthermore, according to the embodiment, when it is determined that the user is not awake even after the call, a call is made to the user to suggest active behavior that is expected to improve the user's alertness even if it takes some time. This gives the user an opportunity to take action, and as a result, the user's alertness is expected to improve in a short period of time. Furthermore, the voice of the call to suggest active behavior is made in a voice that is familiar to the user, which has the effect of making the user hear the voice of the call due to the cocktail party effect.

[Modification]
A modified example of the embodiment will be described. In the embodiment, the call sentence used in the call voice is randomly selected from templates recorded in the call sentence DB 54. This template can be changed as appropriate. For example, by collecting daily conversations of users, the template can be changed to include words that are likely to attract the user's attention and that appear frequently in daily conversations.

In addition, in the embodiment, the voice label selection unit 232 selects a voice label corresponding to the maximum familiarity among the acquired familiarities, or a voice label corresponding to a combination of familiarity, response probability, and average awakening change value that maximizes the weighted sum of these familiarity, response probability, and average awakening change value among the acquired familiarities. Not limited to this, the voice label selection unit 232 may extract, as candidate voice labels for the call voice, a plurality of voice labels corresponding to a plurality of familiarities with the highest magnitude among the acquired familiarities, or a plurality of voice labels corresponding to a combination of a plurality of familiarities, response probability, and average awakening change value with the highest magnitude of these weighted sums among the acquired familiarities, response probability, and average awakening change value. Then, the voice label selection unit 232 may select one of the extracted candidate voice labels by random sampling. This suppresses the user's familiarity and boredom due to the frequent presentation of the call voice with the same voice label. This makes it easier to expect a user's reaction to the call voice even when the voice generation device 1 is used for a long period of time, and as a result, an increase in the user's awakening level is expected.

In addition, in an embodiment, when a voice prompt is generated, changes to the volume, etc. may also be made according to the level of alertness.

In addition, in the embodiment, active behavior is extracted based on the relationship between the level of arousal and the time required. However, instead of or in addition to the time required, the level of interest of the user in the active behavior, the strength of the physical stimulation, etc. may be used to extract active behavior.

In addition, each process according to the above-described embodiment can be stored as a program that can be executed by a processor, which is a computer. Alternatively, the program can be stored in a storage medium of an external storage device, such as a magnetic disk, optical disk, or semiconductor memory, and distributed. The processor can then read the program stored in the storage medium of the external storage device, and have its operation controlled by the read program, thereby executing the above-described process.

Note that the present invention is not limited to the above-described embodiments, and can be modified in various ways in the implementation stage without departing from the gist of the invention. The embodiments may also be implemented in appropriate combination, in which case the combined effects can be obtained. Furthermore, the above-described embodiments include various inventions, and various inventions can be extracted by combinations selected from the multiple constituent elements disclosed. For example, if the problem can be solved and an effect can be obtained even if some constituent elements are deleted from all the constituent elements shown in the embodiments, the configuration from which these constituent elements are deleted can be extracted as an invention.

1... Sound generating device 2... Processor 3... ROM
4...RAM
5...Storage 6...Microphone (Mic)
7: Speaker 8: Camera 9: Input device 10: Display 11: Communication module 21: Acquisition unit 22: Determination unit 23: Selection unit 24: Generation unit 25: Presentation unit 51: Familiarity database (DB)
52...User log database (DB)
53...Speech synthesis parameter database (DB)
54...Call sentence database (DB)
55... Active behavior database (DB)
231: Receiving section 232: Voice label selecting section 233: Transmitting section

Claims (7)

An acquisition unit that acquires a first awakening level that represents a degree of awakening of a user from sleep to excitement;
A determination unit that determines whether the user is in an awake state based on the first wakefulness level;
A selection unit that selects a voice that prompts the user to wake up from among a plurality of voice candidates when the user is not in an awake state;
a generator that generates a first call voice to be presented to the user based on the selected voice;
Equipped with
Each of the plurality of voice candidates is associated with a familiarity level representing a degree of familiarity with the corresponding voice candidate by the user; and
The selection unit selects a sound that prompts the user to wake up in accordance with the familiarity degree.
A voice generating device. Each of the plurality of voice candidates is further associated with a reaction of the user to the presented first call voice and a change amount representing a change in an alertness of the user in response to the presented first call voice,
The voice generating device according to claim 1 , wherein the selection unit selects a voice that encourages the user to wake up, further depending on the reaction and the amount of change. The acquisition unit acquires a second awakening level after the user responds to the presented first call voice,
The determination unit determines again whether or not the user is in an awake state based on the second wakefulness level,
The generation unit generates, when the user is not in an awake state, a second call voice for encouraging the user to take an active action to encourage the user to wake up.
3. A voice generating device according to claim 1 or 2 . The voice generating device according to claim 3 , wherein the generating unit generates the second call voice based on the voice selected by the selecting unit. The voice generation device according to any one of claims 1 to 4, wherein the first level of alertness is calculated based on any one or a combination of the user's eye movement, the user's blink activity, the user's electrical skin activity, and the user's reaction time to a stimulus. acquiring, by an acquisition unit, a first awakening level representing a degree of awakening of the user ranging from sleep to excitement;
determining, by a determination unit, whether or not the user is in an awake state based on the first wakefulness level;
selecting, by a selection unit, a voice that prompts the user to wake up from among a plurality of voice candidates when the user is not in an awakened state;
generating, by a generating unit, a first call voice to be presented to the user based on the selected voice;
Equipped with
Each of the plurality of voice candidates is associated with a familiarity level representing a degree of familiarity with the corresponding voice candidate by the user; and
The selection unit selects a sound that prompts the user to wake up according to the familiarity.
A method for generating audio. A voice generation program for causing a processor to function as the acquisition unit, the determination unit, the selection unit, and the generation unit of the voice generation device according to claim 1 .

Images