Industrial vehicle, alarm sound output direction estimation system, and alarm sound output direction estimation program

Abstract

To provide an industrial vehicle capable of preventing a person from falling due to being surprised by notification sound.SOLUTION: An industrial vehicle includes: a speaker 12 that has directivity and outputs notification sound to a person; a learned model 20 including a direction estimation unit 25; and a direction change unit 26 that changes a notification sound output direction of the speaker 12. The direction estimation unit 25 performs machine learning in advance by using training data using a notification distance from the speaker 12 to the person and a notification sound output direction as input data and a degree of surprise of the notified person as output data. When the notification distance is input, the direction estimation unit estimates a notification sound output direction with which a degree of surprise of the notified person becomes equal to or less than a predetermined degree of surprise. The direction change unit 26 changes the notification sound output direction of the speaker 12 to the estimated notification sound output direction.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to an alert sound output direction estimation system for use in an industrial vehicle, an industrial vehicle equipped with the system, and an alert sound output direction estimation program. [Background technology]

[0002] In recent years, the proportion of workers aged 60 or over among those killed or injured in work-related accidents has been increasing. Work-related accidents involving the elderly include falls while working. Therefore, it is necessary to create a work environment that reduces the number of falls among the elderly.

[0003] Industrial vehicles such as forklifts are equipped with speakers that warn of their approach while in motion, and the warning sounds help to avoid contact with pedestrians (see, for example, Patent Documents 1 and 2). However, many elderly people have hearing difficulties and tend to have difficulty hearing the warning sounds. Therefore, it is conceivable to make the warning sounds louder, but if the warning sounds are too loud, there is a risk that they will startle pedestrians and cause them to fall over. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Publication No. 6-149373 [Patent Document 2] Japanese Patent Application Publication No. 2022-139495 Summary of the Invention [Problem to be solved by the invention]

[0005] Therefore, an object of the present invention is to provide an industrial vehicle that can prevent people from being startled by the alarm sound and falling over. [Means for solving the problem]

[0006] In order to solve the above problems, the industrial vehicle according to the present invention comprises: a speaker having directionality and outputting an alarm sound to a person; a trained model having a direction estimation unit; a direction change unit that changes the direction in which the notification sound is output from the speaker; the direction estimation unit performs machine learning in advance using training data in which the notification distance from the speaker to the person and the notification sound output direction are input data and the degree of surprise of the person who is notified is output data, and when the notification distance is input, the unit estimates the notification sound output direction in which the degree of surprise of the person who is notified is equal to or less than a predetermined level; The direction change unit changes the notification sound output direction of the speaker to the estimated notification sound output direction.

[0007] The industrial vehicle preferably has: Further comprising a camera that photographs a person and generates a person image; The trained model further includes a distance estimation unit, The distance estimation unit performs machine learning in advance using training data in which a person image is input data and the reported distance from the speaker to the person is output data, and when a person image is input, it estimates the reported distance from the speaker to the person.

[0008] The industrial vehicle preferably has: Further comprising a camera that photographs a person and generates a person image; The trained model further includes a posture estimation unit, The posture estimation unit performs machine learning in advance using training data in which a person's image is input data and the type of posture of the person is output data, and when a person's image is input, the unit estimates the type of posture of the person, When the type of posture of the person is input, the direction estimation unit estimates the notification sound output direction in which the surprise level is equal to or less than a predetermined level based on the type of posture of the person.

[0009] The industrial vehicle preferably has: Further comprising a camera that photographs a person and generates a person image; The trained model further includes an orientation estimation unit, The orientation estimation unit performs machine learning in advance using training data in which a person's image is input data and the orientation of the person's face is output data, and when a person's image is input, it estimates the orientation of the person's face, When the direction of the person's face is input, the direction estimation unit estimates the notification sound output direction that will result in a surprise level equal to or lower than a predetermined level based on the direction of the person's face.

[0010] The industrial vehicle preferably has: further comprising an environmental sound acquisition unit that acquires environmental sounds; When the environmental sound is input, the direction estimation unit estimates, based on the environmental sound, an output direction of the notification sound that will produce a surprise level equal to or lower than a predetermined level.

[0011] The industrial vehicle preferably has: Further comprising a camera that photographs a person and generates a person image; The trained model further includes a surprise level estimation unit, the surprise level estimation unit performs machine learning in advance using training data in which the image of the notified person is input data and the surprise level is output data, and when the image of the notified person is input, the surprise level estimation unit estimates the surprise level; The direction estimation unit re-learns the output direction of the notification sound that results in a predetermined surprise level or less based on the image of the person before the notification, the notification distance, the volume of the notification sound when it is given, and the surprise level of the person who is notified.

[0012] The industrial vehicle preferably has: Further provided is an irradiation unit that irradiates light with directionality, The illuminating unit emits light to warn of the approach of an industrial vehicle.

[0013] In order to solve the above problem, the informative sound output direction estimation system according to the present invention comprises: An alarm sound output direction estimation system for use in an industrial vehicle having a directional speaker that outputs an alarm sound to a person, a trained model having a direction estimation unit; a direction change unit that changes the direction in which the notification sound is output from the speaker; the direction estimation unit performs machine learning in advance using training data in which the notification distance from the speaker to the person and the notification sound output direction are input data and the degree of surprise of the person who is notified is output data, and when the notification distance is input, the unit estimates the notification sound output direction in which the degree of surprise of the person who is notified is equal to or less than a predetermined level; The direction change unit changes the notification sound output direction of the speaker to the estimated notification sound output direction.

[0014] In order to solve the above problem, the informative sound output direction estimation program according to the present invention is An informative sound output direction estimation program used in an informative sound output direction estimation system, The notification sound output direction estimation system is a system used in an industrial vehicle having a directional speaker that outputs a notification sound to a person, the system including a computer, The alarm sound output direction estimation program is executed by a computer as a direction estimation unit that performs machine learning in advance using training data in which the alarm distance from the speaker to a person and the alarm sound output direction are input data and the degree of surprise of the person who is notified is output data, and that, when the alarm distance is input, estimates the alarm sound output direction in which the degree of surprise of the person who is notified is below a predetermined level. [Effects of the Invention]

[0015] The industrial vehicle according to the present invention can prevent a person from falling over due to being startled by the alarm sound while still achieving the effect of the alarm sound. [Brief explanation of the drawings]

[0016] [Figure 1] 1 is a schematic diagram showing an industrial vehicle according to an embodiment of the present invention and its surroundings. [Figure 2] FIG. 2 is a block diagram of a control unit of the industrial vehicle. [Figure 3] FIG. 2 is a schematic plan view showing the directivity of a speaker. [Figure 4] 1 is a schematic side view showing the directivity of a speaker, where A shows an example of the direction in which the speaker outputs an alarm sound when a person is far away, and B shows an example of the direction in which the speaker outputs an alarm sound when a person is close. [Figure 5] FIG. 10 is a diagram illustrating the operation of each estimation unit of the trained model. [Figure 6] FIG. 1 is a diagram illustrating an example of a person's posture. [Figure 7] FIG. 10 is a diagram illustrating machine learning of a direction estimation unit. [Figure 8] FIG. 10 is a diagram illustrating the operation of a direction estimation unit. [Figure 9] FIG. 10 is a diagram showing the state where the irradiating unit irradiates light. [Figure 10] FIG. 10 is a flowchart showing an operation of an industrial vehicle to notify a person of its approach. DETAILED DESCRIPTION OF THE INVENTION

[0017] Hereinafter, an embodiment of an industrial vehicle, an alert sound output direction estimation system, and an alert sound output direction estimation program according to the present invention will be described with reference to the accompanying drawings.

[0018] Fig. 1 is a schematic diagram showing an industrial vehicle 1 according to one embodiment of the present invention and its surroundings. As shown in Fig. 1, the industrial vehicle 1 is a forklift truck moving forward along an aisle R. In front of the industrial vehicle 1 is a pedestrian H1 (hereinafter sometimes referred to as "person H"), and next to the aisle R is a worker H2 working.

[0019] The industrial vehicle 1 is an autonomously traveling unmanned guided vehicle, but this is merely an example, and the industrial vehicle according to the present invention may be a manned industrial vehicle. Furthermore, the industrial vehicle 1 does not have to be a forklift, and may be, for example, a forkless transport vehicle, a forklift towing vehicle, a straddle carrier, or the like, and is not particularly limited.

[0020] As shown in Fig. 1, the industrial vehicle 1 includes a vehicle body 10 and a fork 11. The fork 11 is configured to be able to rise and fall, thereby lifting and lowering a load.

[0021] As shown in FIG. 2, the industrial vehicle 1 includes a speaker 12, a camera 13, an irradiation unit 14, and a control unit 18.

[0022] The speaker 12 outputs a directional alarm sound V to the person H. The speaker 12 is made up of a plurality of parametric speakers, and is configured so that the alarm sound output direction DV can be changed in the horizontal and vertical directions, as shown in FIGS. 3 and 4. This prevents the speaker 12 from outputting the alarm sound V to anyone other than the targeted person H. Therefore, it is possible to prevent the alarm sound V from being output to a worker H2 who is not positioned in the traveling direction of the industrial vehicle 1, and thus to prevent the worker H2 from being startled.

[0023] The volume of the informative sound V becomes louder toward the center of the directional width. Therefore, the speaker 12 can change the volume of the informative sound V perceived by the person H by changing the informative sound output direction DV. The speaker 12 is also configured to be able to change the directional width (directional angle) of the informative sound V, which also makes it possible to adjust the volume difference of the informative sound V between directional widths.

[0024] Since the industrial vehicle 1 moves forward and backward, the speakers 12 are configured to be able to output the alarm sound V at least in the forward and backward directions of the industrial vehicle 1. Furthermore, the number of speakers 12 and the configuration of the directionality of the speakers 12 are not particularly limited.

[0025] There are no particular limitations on the specific sound of the notification sound V. For example, the notification sound V may be a buzzer sound, a specific melody, or a human voice.

[0026] The speaker 12 may not output the alarm sound V unless a person H is detected by the industrial vehicle 1. Alternatively, the speaker 12 may be configured to output the alarm sound V at a predetermined volume only at an intersection of the passage R, even if a person H is not detected. In this embodiment, the speaker 12 is configured to output the alarm sound V when a person H is detected by the camera 13 (i.e., when a photograph of the person H is taken).

[0027] The industrial vehicle 1 further includes a melody output unit (not shown) that outputs a melody around the vehicle while the vehicle is traveling. The volume of the melody is set to a volume that will not startle the person H.

[0028] The camera 13 is configured to be able to capture images of the area around the vehicle body 10, captures an image of a person H, and generates a person image HI. The camera 13 is made up of multiple cameras 13, and is configured to be able to capture images of at least the area in front of and behind the industrial vehicle 1.

[0029] The irradiating unit 14 is configured to be able to irradiate directional light L. The light L irradiated from the irradiating unit 14 may be of multiple colors, or may be configured to be able to switch between the multiple colors. The irradiating unit 14 may be configured to irradiate the light L by flashing it. The irradiating units 14 are disposed at the front and rear of the vehicle body 10, and are configured to be able to irradiate the light L at least in the forward and backward directions of the industrial vehicle 1.

[0030] The control unit 18 includes a computer having a storage device, a memory, and an arithmetic unit, and the storage device stores an informative sound output direction estimation program. The informative sound output direction estimation program causes the computer to function as a trained model 20 and a direction changing unit 26, which will be described later.

[0031] <Configuration of trained model> 2, the control unit 18 has, as its functional configuration, a trained model 20 and a direction change unit 26. The trained model 20 has a surprise level estimation unit 21, a distance estimation unit 22, a posture estimation unit 23, a direction estimation unit 24, and a direction estimation unit 25.

[0032] The surprise level estimation unit 21 is, for example, a neural network, and has previously learned the correlation between the image HI of the notified person H as input data and the surprise level Q as output data through training data. The machine learning method may be, for example, a machine learning method that learns the correlation between the difference between the image HI of the notified person H before the notification and the image HI of the notified person H, and the surprise level Q. Alternatively, the surprise level estimation unit 21 may be configured to perform machine learning using facial images expressing emotions such as surprise or impatience as training data, and estimate the surprise level Q of the notified person H when the image HI of the notified person H is input. Through such machine learning, the surprise level estimation unit 21 estimates the surprise level Q of the notified person H when the image HI of the notified person H is input, as shown in FIG. 5A . The surprise level Q may be output as a numerical value between 0.01 and 1.0, for example.

[0033] In the present invention, the "predetermined surprise level RQ" refers to a surprise level Q that can prevent a person from being startled and falling over, dropping something they are holding, or making a mistake in their work. This predetermined surprise level RQ may be a numerical value for the surprise level Q, such as 0.7.

[0034] The distance estimation unit 22 is, for example, a neural network, and has previously learned the correlation between these by machine learning using training data in which the human image HI is input data and the notification distance DN from the speaker 12 to the person H is output data. As a result, when the human image HI is input, the distance estimation unit 22 estimates the notification distance DN from the speaker 12 to the person H, as shown in FIG. 5B . The estimated notification distance DN is output to the direction estimation unit 25. Note that if the human image HI is not input, the distance estimation unit 22 does not output the notification distance DN to the direction estimation unit 25. Note that the "notification distance" in the present invention also includes the direction from the speaker 12 to the person H.

[0035] The posture estimation unit 23 is, for example, a neural network, and has previously learned the correlation between the human image HI and the posture type AP of the person H by machine learning using training data having the human image HI as input data and the posture type AP of the person H as output data. As a result, when the posture estimation unit 23 receives the human image HI as input, it estimates the posture type AP of the person H (hereinafter, may be simply referred to as the "posture AP of the person H") as shown in FIG. 5C. The estimated posture AP of the person H is output to the direction estimation unit 25.

[0036] Fig. 6 is a diagram showing examples of postures AP of person H. As shown in Fig. 6, types of postures AP of person H include, for example, a standing posture AP (see Fig. 6A), a posture AP carrying a load (see Fig. 6B), a walking posture AP (see Fig. 6C), and a working posture AP (see Fig. 6D).

[0037] The orientation estimation unit 24 is, for example, a neural network that has undergone machine learning in advance using training data in which the human image HI is input data and the facial orientation DF of the person H is output data. As a result, when the orientation estimation unit 24 receives the human image HI as input, it estimates the facial orientation DF of the person H, as shown in FIG. 5D . The facial orientation DF of the person H is, for example, the direction of the industrial vehicle 1, the direction opposite to the industrial vehicle 1, the direction of a load that the person H is carrying, or the direction of a work tool when working. The estimated facial orientation DF of the person H is output to the direction estimation unit 25 and the irradiation unit 14.

[0038] The direction estimation unit 25 is, for example, a neural network, and as shown in Fig. 7, the direction estimation unit 25 receives the alert distance DN, the posture AP of the person H, the facial direction DF of the person H, and the alert sound output direction DV as input data, and has previously machine-learned the correlation therebetween using training data that receives the surprise level Q as output data. As a result, when the alert distance DN, the posture AP, and the facial direction DF of the person H estimated by the distance estimation unit 22, the orientation estimation unit 24, and the posture estimation unit 23 are input, the direction estimation unit 25 estimates the alert sound output direction DV that is equal to or less than a predetermined surprise level RQ, as shown in Fig. 8. The estimated alert sound output direction DV is output to the direction change unit 26.

[0039] In this embodiment, the direction estimation unit 25 does not output the notification sound output direction DV to the direction change unit 26 unless the notification distance DN is input (that is, unless the person H is detected).

[0040] In this embodiment, the speaker 12 can change the informative sound output direction DV both horizontally and vertically, so the change direction of the informative sound output direction DV is not limited to either the vertical direction or the horizontal direction. On the other hand, if the speaker 12 is configured to be able to change the informative sound output direction DV only horizontally or vertically, the teacher data for the informative sound output direction DV in the machine learning of the direction estimation unit 25 will also be between horizontal directions or between vertical directions, and the informative sound output direction DV estimated by the direction estimation unit 25 will also be between horizontal directions or between vertical directions. However, as shown in FIG. 1 , there may be another person (e.g., worker H2) next to the aisle R, so the change direction of the informative sound output direction DV (i.e., the direction of the informative sound output direction DV estimated by the direction estimation unit 25) may be limited to between vertical directions.

[0041] The direction change unit 26 changes the alert sound output direction DV of the speaker 12 to the alert sound output direction DV estimated by the direction estimation unit 25. By making the speaker 12 output the alert sound V in the estimated alert sound output direction DV, the direction change unit 26 can prevent the person H from being startled by the alert sound V and falling over.

[0042] Furthermore, when the alarm sound V is actually output, the direction estimation unit 25 uses as training data the alarm distance DN, the posture AP of the person H, the facial orientation DF of the person H, the actually output alarm sound output direction DV, and the surprise degree Q estimated by the surprise degree estimation unit 21 at that time, and re-learns the correlations among them. That is, the direction estimation unit 25 in this embodiment uses as input data the alarm distance DN, the posture AP of the person H, the facial orientation DF, and the actually output alarm sound output direction DV estimated by the distance estimation unit 22, the orientation estimation unit 24, and the posture estimation unit 23, and re-learns the correlations among them, thereby enabling more accurate estimation from the next time.

[0043] The illumination unit 14 illuminates light L in the direction the person H's face is facing based on the facial direction DF estimated by the direction estimation unit 24, thereby alerting the person H to the approach of the industrial vehicle 1. FIG. 9 shows the light L being illuminated toward the load the person H is carrying when the person H is facing the load. In this way, if the alarm sound V is output at a high volume when the person H is facing the load he or she is carrying, the person H may be startled. Therefore, by reducing the volume of the alarm sound V and illuminating light L toward the load, the approach of the industrial vehicle 1 can be alerted without startling the person H. The direction in which the illumination unit 14 illuminates light L may be, for example, the road surface.

[0044] A system including the above-described camera 13, the trained model 20, and the direction change unit 26 corresponds to the informative sound output direction estimation system of the present invention. In this embodiment, the informative sound output direction estimation system further includes an irradiation unit 14.

[0045] <Industrial vehicle operation> Next, the operation of the industrial vehicle 1 will be described again with reference to FIG.

[0046] (1) While the industrial vehicle 1 is traveling, the melody output unit outputs a melody (see S1 (step 1) in FIG. 10). The industrial vehicle 1 sets the volume of the alarm sound V to zero until the person H is photographed by the camera 13 (see S2 in FIG. 10).

[0047] (2) When the industrial vehicle 1 photographs a person H using the camera 13 (i.e., detects a person H) (Yes in S3 of Figure 10), the learned model 20 estimates the alarm distance DN, the posture AP of the person H, and the facial direction DF based on the person image HI (see S4 of Figure 10).

[0048] (3) Next, the industrial vehicle 1 uses the trained model 20 to estimate the alarm sound output direction DV that will result in a level of surprise RQ or less based on the estimated alarm distance DN, the posture AP of the person H, and the facial direction DF (see S5 in Figure 10).

[0049] (4) Next, the industrial vehicle 1 changes the notification sound output direction DV of the speaker 12 by the direction change unit 26 so that the notification sound output direction DV becomes the estimated notification sound output direction DV (see S6 in FIG. 10).

[0050] (5) Next, the industrial vehicle 1 outputs the notification sound V from the speaker 12 and irradiates the light L from the irradiating unit 14 (see S7 in FIG. 10).

[0051] While the industrial vehicle 1 is capturing an image of the person H with the camera 13 (i.e., while the person H is continuously being detected), steps S4 to S7 in Fig. 10 are repeated, and the notification sound output direction DV is automatically changed as the notification distance DN, the posture AP of the person H, and the facial direction DF change. Then, when the camera 13 no longer captures an image of the person H (i.e., when the person H is no longer detected) (No in S3 in Fig. 10), the volume of the notification sound V is changed to zero again (see S2 in Fig. 10).

[0052] The industrial vehicle 1, having the above configuration, can prevent the person H from being startled by the alarm sound V and falling over. Furthermore, the industrial vehicle 1 uses the trained model 20 to learn not only the correlation between the alarm distance DN, the alarm sound output direction DV, and the surprise level Q, but also the correlation between the posture AP and facial direction DF of the person H, so it is possible to prevent the alarm sound V from being output too loud to a person H who is concentrating on work or looking in the opposite direction from the industrial vehicle 1, startling the person H and causing them to fall over. Moreover, because the industrial vehicle 1 estimates the alarm distance DN, the posture AP, and facial direction DF of the person H based on the person image HI, there is no need to add any special sensors.

[0053] Although one embodiment of the industrial vehicle, the alert sound output direction estimation system, and the alert sound output direction estimation program according to the present invention has been described above, the present invention is not limited to the above embodiment. For example, the industrial vehicle, the alert sound output direction estimation system, and the alert sound output direction estimation program according to the present invention may be implemented by appropriately combining the following modified examples.

[0054] <Modification> The industrial vehicle 1 may be equipped with a known sensor capable of detecting the direction of the person H and the notification distance DN. In this case, the direction estimation unit 25 acquires the notification distance DN from the sensor, and the trained model 20 does not need to include the distance estimation unit 22.

[0055] The human image HI used in machine learning of each component of the trained model 20 and the human image HI input when making each estimation may be a plurality of human images HI. In other words, the human image HI according to the present invention is a concept that includes moving images.

[0056] The irradiating unit 14 may irradiate the light L not in the direction of the face DF of the person H, but, for example, toward the ground in the direction in which the person H is facing.

[0057] The input data of the teacher data used in machine learning by the direction estimation unit 25 only needs to include at least the alarm distance DN and the alarm sound output direction DV. In this case, when the alarm distance DN is input, the direction estimation unit 25 estimates the alarm sound output direction DV that is equal to or less than a predetermined surprise level RQ.

[0058] The industrial vehicle 1 may further include an environmental sound acquisition unit that acquires environmental sound. In this case, the direction estimation unit 25 further receives the environmental sound, the alarm distance DN, the posture AP of the person H, the facial orientation DF of the person H, and the alarm sound output direction DV as input data, and performs machine learning to determine correlations among them in advance using training data that outputs a surprise level Q. When the environmental sound, the alarm distance DN, the posture AP, and the facial orientation DF of the person H are input, the direction estimation unit 25 estimates the alarm sound output direction DV that will be equal to or less than a predetermined surprise level RQ. In this case, the direction estimation unit 25 may also receive the environmental sound, the alarm distance DN, and the alarm sound output direction DV as input data, and perform machine learning to determine correlations among them in advance using training data that outputs a surprise level Q. When the environmental sound and the alarm distance DN are input, the direction estimation unit 25 may also estimate the alarm sound output direction DV that will be equal to or less than a predetermined surprise level RQ.

[0059] The alert sound output direction estimation system may further include a server computer capable of communicating with the industrial vehicle 1, and the trained model 20 may be configured by the server computer. In this case, the industrial vehicle 1 communicates with the server computer to receive the alert sound output direction DV that results in a predetermined surprise level RQ or less, and outputs the alert sound V in this alert sound output direction DV. [Explanation of symbols]

[0060] 1. Industrial vehicles (forklifts) 10. Body 11. Fork 12 speakers 13 Camera 14 Irradiation unit 18 Control Unit 20 Pre-trained models 21 Surprise Level Estimation Unit 22 Distance estimation unit 23 Posture estimation section 24 Orientation estimation unit 25 Direction estimation part 26 Direction change section Types of AP posture DF Face Direction DN broadcast distance DV alarm sound output direction H people (pedestrians, workers) HI People Images L light Q Surprise level R aisle RQ Predetermined surprise level V Notification sound

Claims

1. a speaker having directionality and outputting an alarm sound to a person; a trained model having a direction estimation unit; a direction change unit that changes the direction in which the notification sound is output from the speaker, the direction estimation unit performs machine learning in advance using teacher data in which the notification distance from the speaker to the person and the notification sound output direction are input data and the degree of surprise of the person who has been notified is output data, and when the notification distance is input, estimates the notification sound output direction in which the degree of surprise of the person who has been notified is equal to or less than a predetermined value; The direction changer changes the output direction of the notification sound from the speaker to the estimated output direction of the notification sound.

2. a camera for capturing an image of the person and generating a person image; The trained model further includes a distance estimation unit, 2. The industrial vehicle according to claim 1, wherein the distance estimation unit performs machine learning in advance using teacher data in which the human image is input data and the reported distance from the speaker to the person is output data, and when the human image is input, the distance estimation unit estimates the reported distance from the speaker to the person.

3. a camera for capturing an image of the person and generating a person image; The trained model further includes a posture estimation unit, the posture estimation unit performs machine learning in advance using training data in which the human image is input data and a type of posture of the human is output data, and when the human image is input, estimates the type of posture of the human; 2. The industrial vehicle according to claim 1, wherein the direction estimation unit, when receiving input of the type of posture of the person, estimates the output direction of the alarm sound such that the surprise level is equal to or less than a predetermined level based on the type of posture of the person.

4. a camera for capturing an image of the person and generating a person image; The trained model further includes an orientation estimation unit, the orientation estimation unit performs machine learning in advance using training data in which the human image is input data and the orientation of the face of the person is output data, and when the human image is input, estimates the orientation of the face of the person; The industrial vehicle according to claim 1 , wherein the direction estimation unit, upon receiving the direction of the person's face, estimates the output direction of the alarm sound such that the degree of surprise is equal to or less than a predetermined level based on the direction of the person's face.

5. further comprising an environmental sound acquisition unit that acquires environmental sounds; The industrial vehicle according to any one of claims 1 to 4, wherein when the environmental sound is input, the direction estimation unit estimates an output direction of the notification sound that is equal to or less than a predetermined surprise level based on the environmental sound.

6. a camera for capturing an image of the person and generating a person image; The trained model further includes a surprise level estimation unit, the surprise level estimation unit performs machine learning in advance using teacher data in which the human image of the notified person is input data and a surprise level is output data, and when the human image of the person notified by the alarm sound is input, the surprise level estimation unit estimates the surprise level; 2. The industrial vehicle according to claim 1, wherein the direction estimation unit re-learns the output direction of the alarm sound that is below a predetermined surprise level based on the image of the person before the alarm, the alarm distance, the volume of the alarm sound when it is issued, and the surprise level of the person who is notified.

7. Further provided is an irradiation unit that irradiates light with directionality, The industrial vehicle according to claim 1 , wherein the illuminating unit irradiates the light to warn of the approach of the industrial vehicle.

8. An alarm sound output direction estimation system for use in an industrial vehicle having a directional speaker that outputs an alarm sound to a person, a trained model having a direction estimation unit; a direction change unit that changes the direction in which the notification sound is output from the speaker, the direction estimation unit performs machine learning in advance using teacher data in which the notification distance from the speaker to the person and the notification sound output direction of the speaker are input data and the degree of surprise of the person who has been notified is output data, and when the notification distance is input, estimates the notification sound output direction in which the degree of surprise of the person who has been notified is equal to or less than a predetermined value; The direction change unit changes the output direction of the speaker of the informative sound to the estimated output direction of the informative sound.

9. An informative sound output direction estimation program used in an informative sound output direction estimation system, The notification sound output direction estimation system is a system used in an industrial vehicle having a directional speaker that outputs a notification sound to a person, the system including a computer, The informative sound output direction estimation program is configured to: an alarm sound output direction estimation program that performs machine learning in advance using teacher data in which the alarm distance from the speaker to the person and the alarm sound output direction of the speaker are input data and the degree of surprise of the person who has been notified is output data, and that, when the alarm distance is input, estimates the alarm sound output direction in which the degree of surprise of the person who has been notified is equal to or less than a predetermined value.

Images