Classifications

• • B—PERFORMING OPERATIONS; TRANSPORTING
  • B60—VEHICLES IN GENERAL
  • B60R—VEHICLES, VEHICLE FITTINGS, OR VEHICLE PARTS, NOT OTHERWISE PROVIDED FOR
  • B60R11/00—Arrangements for holding or mounting articles, not otherwise provided for
  • B60R11/02—Arrangements for holding or mounting articles, not otherwise provided for for radio sets, television sets, telephones, or the like; Arrangement of controls thereof
• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10K—SOUND-PRODUCING DEVICES; METHODS OR DEVICES FOR PROTECTING AGAINST, OR FOR DAMPING, NOISE OR OTHER ACOUSTIC WAVES IN GENERAL; ACOUSTICS NOT OTHERWISE PROVIDED FOR
  • G10K11/00—Methods or devices for transmitting, conducting or directing sound in general; Methods or devices for protecting against, or for damping, noise or other acoustic waves in general
  • G10K11/16—Methods or devices for protecting against, or for damping, noise or other acoustic waves in general
  • G10K11/175—Methods or devices for protecting against, or for damping, noise or other acoustic waves in general using interference effects; Masking sound
  • G10K11/178—Methods or devices for protecting against, or for damping, noise or other acoustic waves in general using interference effects; Masking sound by electro-acoustically regenerating the original acoustic waves in anti-phase
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; ELECTRIC HEARING AIDS; PUBLIC ADDRESS SYSTEMS
  • H04R1/00—Details of transducers, loudspeakers or microphones
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; ELECTRIC HEARING AIDS; PUBLIC ADDRESS SYSTEMS
  • H04R1/00—Details of transducers, loudspeakers or microphones
  • H04R1/02—Casings; Cabinets ; Supports therefor; Mountings therein
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; ELECTRIC HEARING AIDS; PUBLIC ADDRESS SYSTEMS
  • H04R3/00—Circuits for transducers
  • H04R3/04—Circuits for transducers for correcting frequency response
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; ELECTRIC HEARING AIDS; PUBLIC ADDRESS SYSTEMS
  • H04R7/00—Diaphragms for electromechanical transducers; Cones
  • H04R7/02—Diaphragms for electromechanical transducers; Cones characterised by the construction
  • H04R7/04—Plane diaphragms

Definitions

• This disclosure relates to an information processing device that controls a glass vibrating plate with a vibrator.
• JP 2021-180486 A discloses an example of generating a specific sound by vibrating interior materials or vehicle glass windows, and as one example, discloses a configuration in which one or more sound generators are placed on the front glass window to generate a specific sound.
• the present disclosure aims to provide an information processing device and information processing program that improves the acoustic characteristics in a space surrounded by a housing that includes a glass diaphragm with a vibrator by controlling the glass diaphragm with a vibrator.
• the information processing device includes an estimation unit that estimates the control parameters at each time point from newly acquired input data using an estimation model that estimates the control parameters from the input data, the estimation model being generated by machine learning using teacher data in which control parameters that realize vibration of the oscillator such that the acoustic characteristics of the space obtained by vibration of the glass plate structure approach target acoustic characteristics are associated with input data including a spatial occupancy state that represents the state of a space separated from the surroundings by a housing including a glass vibration plate with an oscillator attached to a glass plate structure with an adjustable opening degree, and the opening degree of the glass plate structure, and a control unit that controls the oscillator according to the control parameters estimated by the estimation unit.
• the information processing device includes an estimation unit that estimates the characteristic parameters at each point from newly acquired input data using an estimation model that estimates the characteristic parameters from the input data, the input data including moving body information representing the moving state of the moving body at a point where the moving body is located and environmental information representing the environment at the point where the moving body is located, the moving body information being generated by machine learning using teacher data in which characteristic parameters representing the characteristics of environmental sounds flowing from outside the moving body into the inside of the moving body at the point where the moving body is located are associated with the characteristic parameters, and a control unit that controls the vibrator so as to attenuate the environmental sounds flowing into the inside of the moving body represented by the characteristic parameters estimated by the estimation unit.
• the information processing device and information processing program disclosed herein improve the acoustic characteristics in a space surrounded by a housing that includes a glass diaphragm with a vibrator by controlling the glass diaphragm with a vibrator.
• FIG. 2 is a cross-sectional view showing an example of a cross section of a glass diaphragm with a vibrator as viewed from the side.
• 1 is a plan view of a vehicle showing an example of a location where a glass vibrating plate with a vibrator is applied to a vehicle.
• FIG. FIG. 2 is a diagram illustrating an example of a front side window.
• 1A and 1B are diagrams illustrating an example of attaching a transducer to a glass plate structure.
• 1 is a schematic diagram of a vehicle in which a glass vibration plate with a vibrator is applied to a front side window and a rear side window.
• FIG. 2 is a diagram illustrating an example of the configuration of a management system for a control unit.
• FIG. 1 is a schematic diagram of a vehicle in which a glass vibration plate with a vibrator is applied to a front side window and a rear side window.
• FIG. 2 is a diagram illustrating an example of the configuration of
• FIG. 2 is a diagram illustrating an example of a functional configuration of a control unit.
• FIG. 13 is a diagram illustrating an example of machine learning of an estimation model.
• FIG. 1 is a schematic diagram illustrating an example of a reflected sound generated inside a vehicle.
• FIG. 2 is a diagram illustrating an example of the configuration of a main part of an electrical system of a control unit.
• 10 is a flowchart showing an example of a flow of a control process for a transducer.
• FIG. 1 is a diagram showing an example of a glass vibrating plate with vibrators in which a plurality of vibrators are attached to one and the same glass plate structure.
• FIG. 11 is a diagram showing an example of head position information.
• 13 is a flowchart showing an example of the flow of an initial process.
• 10 is a flowchart showing an example of a flow of a control process for a transducer.
• 10 is a flowchart showing an example of the flow of an estimation process for estimating a correction amount of a control parameter.
• 10 is a flowchart showing an example of the flow of a noise canceling process.
• 1 is a schematic diagram visualizing a state in which a target sound is output from a glass plate structure 9.
• FIG. 10 is a flowchart showing an example of the flow of an environmental sound attenuation process.
• FIG. 2 is a diagram illustrating an example of a functional configuration of a control unit corresponding to generation of teacher data.
• 13 is a flowchart showing an example of the flow of a teacher data generation process.
• 10 is a flowchart showing an example of the flow of an environmental sound attenuation process.
• Fig. 1 is a cross-sectional view showing an example of a cross section seen from the side of a glass vibrator with a vibrator 1.
• the glass vibrator with a vibrator 1 of this embodiment is configured to include a glass vibrator plate 2 and a vibrator 3, and the vibrator 3 is attached to the glass vibrator plate 2.
• the glass diaphragm 2 includes a glass plate construct 9.
• the glass plate construct 9 of the present embodiment may be made of a single glass plate, but is preferably made of laminated glass from the viewpoint of improving the acoustic effect of the glass diaphragm 2.
• the glass plate structure 9 is formed of transparent or translucent inorganic glass.
• the glass plate structure 9 may be formed of organic glass.
• organic glass include PMMA (polymethyl methacrylate)-based resins, PC (polycarbonate)-based resins, PS (polystyrene)-based resins, PET (polyethyleneterephthalate)-based resins, PVC (polyvinyl chloride)-based resins, and cellulose-based resins.
• the glass plate may be either unreinforced glass or reinforced glass. Unreinforced glass is formed by forming molten glass into a plate shape and slowly cooling it.
• Reinforced glass is formed by forming a compressive stress layer on the surface of unreinforced glass, and may be either air-cooled reinforced glass or chemically reinforced glass.
• the tempered glass is physically tempered glass (e.g., air-cooled tempered glass)
• the glass surface may be strengthened by generating a compressive stress layer on the glass surface due to the temperature difference between the glass surface and the inside of the glass by an operation other than slow cooling, such as rapidly cooling the glass plate uniformly heated in bending from a temperature near the softening point.
• the tempered glass is chemically tempered glass, the glass surface may be strengthened by generating a compressive stress on the glass surface by an ion exchange method or the like after bending.
• a glass plate that absorbs ultraviolet or infrared rays may be used as the glass plate construct 9.
• the glass plate construct 9 is preferably transparent, but may be a glass plate that is colored to the extent that transparency is not impaired.
• the glass plate construct 9 When the glass plate construct 9 is formed by laminated glass including a plurality of glass plates, an intermediate layer may be sandwiched between a pair of glass plates, but a configuration having three or more glass plates may also be used.
• the glass plate construct 9 may have a curved shape such that the exterior side is convex when attached to the vehicle C. When the glass plate construct 9 has a curved shape, each of the pair of glass plates is bent using gravity forming, press forming, roller forming, or the like.
• the pair of glass plates may have the same thickness or different thicknesses.
• the thickness of the glass plate located outside the vehicle is preferably 1.0 mm or more and 3.0 mm or less.
• the thickness of the glass plate located outside the vehicle is 1.0 mm or more, the strength of the stone chipping resistance performance and the like is sufficient, and if it is 3.0 mm or less, the mass of the glass plate construct 9 is not too large, which is preferable in terms of fuel efficiency of the vehicle C.
• the thickness of the glass plate located on the inside of the vehicle is preferably 0.3 mm or more and 2.3 mm or less.
• the glass plate has good handleability, and when the thickness is 2.3 mm or less, the mass does not become too large.
• the glass plate structure 9 can be both lightweight and soundproof, which is preferable.
• the glass plate located on the inside of the vehicle may be chemically strengthened glass.
• the compressive stress value of the glass surface is preferably 300 MPa or more, and the depth of the compressive stress layer is preferably 2 ⁇ m or more.
• the thickness of the laminated glass is preferably 1.0 mm or more, more preferably 2.0 mm or more, and even more preferably 3.0 mm or more. This allows the laminated glass to have a necessary and sufficient strength.
• each glass plate constituting the laminated glass is preferably 5.0 mm or less, more preferably 3.0 mm or less, and even more preferably 2.0 mm or less. Furthermore, the thickness of each glass plate constituting the laminated glass is preferably 0.1 mm or more, more preferably 0.5 mm or more, and even more preferably 1.0 mm or more. The thicknesses of the pair of glass plates may be the same or different.
• the intermediate layer that constitutes the laminated glass is formed from a transparent resin film such as polyvinyl butyral (PVB)-based, ethylene-vinyl acetate copolymer (EVA)-based, silicone (PDMS)-based, polyurethane (TPU)-based, fluorine-based, polyethylene terephthalate-based, polycarbonate-based, etc.
• a transparent resin film such as polyvinyl butyral (PVB)-based, ethylene-vinyl acetate copolymer (EVA)-based, silicone (PDMS)-based, polyurethane (TPU)-based, fluorine-based, polyethylene terephthalate-based, polycarbonate-based, etc.
• PVB polyvinyl butyral
• EVA ethylene-vinyl acetate copolymer
• PDMS silicone
• TPU polyurethane
• fluorine-based polyethylene terephthalate-based
• polycarbonate-based etc.
• the intermediate layer that constitutes the laminated glass may be formed using a resin that contains one or more of the components PVB, EVA, TPU, and PDMS, and that contains a material with a dimming function that can change color by passing an electric current through it.
• the intermediate layer is not limited to the above-mentioned resin film, but may also be a gel layer, an adhesive layer, a liquid layer, a sol layer, or a grease layer.
• the thickness of the intermediate layer may be set to, for example, 0.1 [ ⁇ m] to 3.0 [mm], 1 [ ⁇ m] to 2.8 [mm], or 3.0 [ ⁇ m] to 2.6 [mm].
• the mount 7 is fixed to one of the main surfaces of the glass plate construct 9 via a resin layer 8.
• the direction from the glass plate construct 9 toward the mount 7 is referred to as the "upward direction,” and the opposite direction is referred to as the "downward direction.”
• the up-down direction referred to here may be a direction different from the up-down direction in a state in which the glass diaphragm 2 is assembled to a frame or the like.
• the mount 7 is not essential, and the vibrator 3 may be attached to one of the main surfaces of the glass plate construct 9 without the mount 7.
• the resin layer 8 has the same outer diameter as the mounting portion 7, and is provided over the entire lower surface of the mounting portion 7.
• an adhesive a pressure-sensitive adhesive, or the like can be appropriately used.
• a sheet-shaped adhesive tape can be used as the pressure-sensitive adhesive.
• the resin layer 8 may be configured to include an acrylic resin adhesive, but is not limited to this. Furthermore, the mount portion 7 and the glass plate structure 9 may be fixed mechanically.
• connection portion 6 is provided on the side of the mount portion 7 opposite to the glass plate construct 9.
• the glass plate construct 9 is disposed on the lower surface of the mount portion 7, and the connection portion 6 is disposed on the upper surface of the mount portion 7.
• connection part 6 may, as an example, form the outer shell of the vibrator 3.
• the vibrator 3 may be assembled with its bottom surface open, and the open bottom surface may be closed by the connection part 6.
• a part of the connection part 6 may be configured as a lid that covers a part of the vibrator 3.
• the vibrator 3 may be attached to the connection part 6 mechanically with a screw, bolt, key, pin, etc., or may be attached to the connection part 6 with an adhesive, etc.
• the vibrator 3 is connected to a power source (not shown) and vibrates the glass plate construct 9 according to the magnitude of the input voltage.
• the vibrator 3 in this embodiment is a voice coil motor including a coil portion and a magnetic circuit, one of the coil portion and the magnetic circuit is fixed to the mount portion 7, and the other is arranged so as to be movable relative to the mount portion 7.
• vibration is generated by the interaction between the coil portion and the magnetic circuit, and the glass plate construct 9 is vibrated via the mount portion 7.
• the vibrator 3 is not limited to a voice coil motor, and may be an actuator other than a voice coil motor, such as a piezoelectric actuator, as long as it is an actuator capable of transmitting a desired vibration to the glass plate construct 9.
• the glass plate component 9 of the glass diaphragm with vibrator 1 generally has a higher sound reflectivity than interior wall materials that make up a space, such as sound-absorbing boards and wallpaper. Therefore, for example, when a person is listening to music or other sounds in a space (hereinafter simply referred to as "space") surrounded by a housing that includes the glass diaphragm with vibrator 1, the person's ears will hear a mixture of sounds that reach the person directly from a sound source such as a speaker without being reflected by the glass plate component 9 of the glass diaphragm with vibrator 1, and sounds that reach the person after being reflected by the glass plate component 9 of the glass diaphragm with vibrator 1.
• space a space surrounded by a housing that includes the glass diaphragm with vibrator 1
• the person's ears will hear a mixture of sounds that reach the person directly from a sound source such as a speaker without being reflected by the glass plate component 9 of the glass diaphragm with vibrator 1, and sounds that reach the
• the sounds that reach the person's ears after being reflected by the glass plate component 9 of the glass diaphragm with vibrator 1 may also differ in the number of times each sound is reflected, making the sounds difficult to hear due to reverberation or sounds that are not to the person's liking.
• the transducer 3 so as to attenuate the reflected sound generated when the sound that the person wishes to hear (hereinafter referred to as the "target sound") is reflected off the glass plate structure 9, the reflected sound reflected from the glass plate structure 9 is reduced, making it easier for the person to hear the direct sound reproduced from a speaker or the like. Furthermore, by adjusting the magnitude and phase of the reflected sound by controlling the vibration amount and phase of the transducer 3, the sound of the sound source can be made to approach the person's preferred sound without adjusting the frequency characteristics of the sound itself output from the sound source.
• the vibrator 3 if a control signal corresponding to an emphasis sound whose phase has been adjusted to be in phase with the reflected sound is input to the vibrator 3 so as to emphasize the reflected sound generated when the target sound is reflected by the glass plate component 9, the sound pressure of the reflected sound reflected by the glass plate component 9 will increase compared to when the glass plate component 9 is not vibrated, making it easier for people to hear the direct sound reproduced from a speaker or the like. Also, by adjusting the magnitude and phase of the emphasis sound by controlling the vibration amount and phase of the vibrator 3, the sound of the sound source can be made to sound closer to the desired sound.
• the glass diaphragm with vibrator 1 has these characteristics, it is applied to devices in a variety of locations.
• the glass diaphragm with vibrator 1 is applied to spaces where people may enter. Specifically, it is applied to moving objects such as vehicles, aircraft, helicopters, and ships, as well as buildings and structures such as homes and soundproof rooms.
• moving objects such as vehicles, aircraft, helicopters, and ships
• buildings and structures such as homes and soundproof rooms.
• the control of the glass diaphragm with vibrator 1 will be explained using the example of application to the window glass of a vehicle.
• FIG. 2 is a plan view of vehicle C showing an example of an application location of the vibrator-equipped glass vibration plate 1 on vehicle C.
• the vibrator-equipped glass vibration plate 1 of the present disclosure is applied to window glass that opens when operated by a person. Therefore, the vibrator-equipped glass vibration plate 1 is applied to the front side window FSW and the rear side window RSW. If the roof glazing RG has a structure that allows it to be opened, the vibrator-equipped glass vibration plate 1 is also applied to the roof glazing RG.
• FIG 3 shows an example of a front side window FSW.
• the vibrator-equipped glass vibration plate 1 is supported by an enclosing member 15 and attached to the door of the vehicle C.
• the front side window FSW is opened and closed by the glass plate component 9 sliding in the direction of gravity, i.e., the vertical direction of the vehicle C, within the window frame 15A that surrounds the periphery of the glass plate component 9.
• the imaginary straight line that runs along the bottom edge of the window frame 15A is referred to as the "belt line BL.”
• the opening 4 The area within the window frame 15A where the glass plate component 9 is no longer present due to the glass plate component 9 sliding downward is called the opening 4.
• the ratio of the opening 4 to the area surrounded by the window frame 15A is called the opening degree.
• the opening degree represents the degree to which the glass plate component 9 is open in the window frame 15A.
• An opening degree of 0% represents a fully closed state in which the glass plate component 9 covers all of the area surrounded by the window frame 15A.
• An opening degree of 100% represents a fully open state in which the glass plate component 9 is not present in any of the areas surrounded by the window frame 15A.
• the glass plate structure 9 is opened and closed, for example, by sliding along the traveling direction of the vehicle C.
• the vibrator-equipped glass vibration plate 1 is not necessarily applicable only to window glass that opens.
• the vibrator-equipped glass vibration plate 1 can also be applied to window glass that does not have an opening structure, but in this embodiment, an example in which the vibrator-equipped glass vibration plate 1 is used for an opening window glass is described.
• window glass that does not have an opening structure include a wind deflector installed in an open-top car or the like to reduce wind entering the car interior, the windshield WS in FIG. 2, the front quarter window FQW located in front of the front side window FSW, adjacent to the A-pillar AP, and the rear window RW, which are provided to ensure the driver's forward diagonal visibility.
• examples of window glass that does not have an opening structure include various quarter windows that are adjacent to pillars such as the B-pillar and C-pillar, and are provided to ensure the driver's rear diagonal visibility.
• the opening degree of the glass plate structure 9 can be treated as 0%.
• Figure 4 shows an example of attaching the vibrator 3 to the glass plate structure 9.
• the vibrator 3 is attached to an area A1 that is below a line located on the beltline BL of the vehicle C when the glass plate structure 9 is in a fully closed state within the window frame 15A.
• the beltline BL is an imaginary line that corresponds to the lower side of an area A2 that is located in an area surrounded by the window frame 15A when the glass plate structure 9 is in a fully closed state within the window frame 15A.
• the vibrator 3 of the present disclosure is attached to a position that is not visible to a person riding in the vehicle C (hereinafter referred to as an "occupant").
• the vibrator 3 may be attached to a position that is visible to an occupant when the glass plate structure 9 is in a fully closed state within the window frame 15A.
• the occupant is an example of a person in space.
• FIG. 5 is a schematic diagram of a vehicle C in which the vibrator-equipped glass vibrating plate 1 is applied to the front side window FSW and the rear side window RSW.
• vehicle C includes a glass diaphragm with a vibrator 1, a speaker 5, a control unit 11, a group of sensors 12, a seat 14, a microphone 16, and a camera 17.
• the speaker 5 is connected to, for example, a car navigation device (not shown) attached to the vehicle C, and outputs route guidance voice, music, radio, and television audio from the car navigation device.
• the speaker 5 can also output audio from a smartphone connected to the car navigation device via short-range wireless communication such as Bluetooth (registered trademark).
• short-range wireless communication such as Bluetooth (registered trademark).
• the microphone 16 collects sounds inside the vehicle. There are no restrictions on the position or number of microphones 16 installed inside the vehicle. The positions of the microphones 16 are fixed in advance, but the positions of the microphones 16 may be changed as necessary. For example, in order to more accurately collect sounds outside the vehicle that flow into the vehicle from the glass plate structure 9, it is preferable to attach the microphones 16 to the glass plate structure 9.
• Camera 17 captures the state inside the vehicle. There are no restrictions on the location or number of cameras 17 installed inside the vehicle. The location of camera 17 is fixed in advance, but the location of camera 17 may be changed as necessary. Note that the image captured by camera 17 may be a video-like image or a still image. In this disclosure, still images and videos will be collectively referred to as "images.”
• the image captured by the camera 17 includes attributes that represent, for example, the spatial occupancy state inside the vehicle. Attributes that represent the spatial occupancy state indicate the state of the space by focusing on objects that exist in a space such as the inside of a vehicle, and are information that represents, for example, what objects are present where in the vehicle and in what state. Attributes that represent the spatial occupancy state include at least one of attributes related to the seats 14 of the vehicle C and attributes related to the baggage 13 brought into the vehicle C, as well as the positions and number of occupants in the vehicle C.
• the attributes related to the seat 14 of the vehicle C include at least one of the shape of the seat 14, the position of the seat 14, and the material of the seat 14.
• the seat 14 of the vehicle C is often equipped with a reclining function, and the angle of the backrest relative to the seat surface can be adjusted. Therefore, the shape of the seat 14 in this disclosure refers to the shape of the seat 14 that changes by adjusting the angle of the backrest. In other words, the shape of the seat 14 is represented by the inclination angle ⁇ of the backrest relative to the seat surface.
• the installation position of the seat 14 can be adjusted by a function for adjusting the seat position along the traveling direction of vehicle C (also referred to as a function for adjusting the fore-aft position of the seat 14) and a function for adjusting the seat position along the direction of gravity (also referred to as a function for adjusting the up-down position of the seat 14). Therefore, the fore-aft and up-down positions of the seat 14 are also examples of attributes related to the seat 14 of vehicle C.
• the material used for the seats 14, such as genuine leather or fabric, varies depending on the vehicle model and grade of the vehicle C. Since the material of the seats 14 is also an element that indicates the state of the interior of the vehicle, it is included in the attributes related to the seats 14 of the vehicle C.
• the attributes related to the luggage 13 include at least one of the position of the luggage 13, the shape of the luggage 13, and the material of the luggage 13.
• the material of the luggage 13 refers to the material that constitutes the external shape of the luggage 13. Therefore, if the luggage 13 is packed in cardboard or the like, the material of the luggage 13 will be cardboard. Also, if the luggage 13 is a bag, the material of the luggage 13 will be, for example, cloth.
• the material of the luggage 13 can be inferred by identifying the type of luggage 13 from an image captured by the camera 17, for example. For example, if the luggage 13 has a box-like shape and a color commonly used for cardboard, the material of the luggage 13 is inferred to be cardboard. In other words, if the material is associated with the type of luggage 13 in advance, the material of the luggage 13 can be inferred once the type of luggage 13 can be identified.
• the positions and number of occupants in vehicle C are obtained from the image captured by the camera 17 of the present disclosure. Furthermore, if luggage 13 is brought into the vehicle, the position, shape, and material of the luggage 13 are obtained from the image captured by the camera 17 of the present disclosure.
• the inclination angle ⁇ of the backrest of the seat 14 and the position of the seat 14 can also be obtained from the captured image by comparing a reference image, which is an image of the seat 14 with the inclination angle ⁇ of the backrest of the seat 14 and the position of the seat 14 adjusted to known values, with an image captured under the same shooting conditions as the reference image by a camera 17 in the same position as the camera 17 that captured the reference image; however, in this embodiment, the inclination angle ⁇ of the backrest of the seat 14 and the position of the seat 14 are obtained from the values of the sensor group 12.
• the sensor group 12 is composed of various sensors for identifying the driving state of the vehicle C.
• the sensor group 12 includes sensors for acquiring information on the vehicle C itself, such as the speed, acceleration, inclination, vibration, direction of travel, and position of the vehicle C, the opening degree of the glass plate components 9 in the front side window FSW and rear side window RSW, the operation amount of the steering wheel, brake pedal, and accelerator pedal, the inclination angle ⁇ of the seat 14, and the position of the seat 14, as well as information on the environment in which the vehicle C is driving, such as signs, illuminance, and raindrops.
• the sensor group 12 also includes various sensors for detecting the positions of pedestrians and other vehicles around the vehicle, and for implementing driving assistance such as driving by following the vehicle C in front.
• the driving state of the vehicle C is an example of the moving state of a moving body.
• the control unit 11 is connected to, for example, the vibrator 3 of the glass vibrating plate 1 with a vibrator, the sensor group 12, the microphone 16, and the camera 17. Using information obtained from the sensor group 12, the microphone 16, and the camera 17, the control unit 11 estimates control parameters for the vibrator 3 such that the acoustic characteristics at each point in time inside the vehicle obtained by the vibration of the glass plate structure 9 approach the target acoustic characteristics, and controls the vibrator 3 according to the estimated control parameters.
• control parameter is a value that specifies a control signal for controlling the vibration of the vibrator 3, such as the vibration amount or phase of the vibrator 3.
• the vibration amount of the vibrator 3 may be expressed by the vibration amount for each frequency, i.e., the frequency distribution of the vibration amount that indicates the magnitude of vibration for each frequency.
• FIG. 6 is a diagram showing an example of the configuration of a management system 10 for a control unit 11 that controls the vibrator 3 of the glass vibrating plate 1 with a vibrator.
• each vehicle C having a control unit 11 is connected to the Internet 18 to which a management server 20 is connected.
• the management server 20 is an example of an external device that manages the control unit 11 provided in each vehicle C.
• the management server 20 includes a learning unit 20A, which uses machine learning to generate an estimation model M that the control unit 11 uses to estimate the control parameters of the vibrator 3.
• the management server 20 distributes the estimation model M to each vehicle C by transmitting the estimation model M generated by the management server 20 to each vehicle C via the Internet 18. Details of the estimation model M will be explained later.
• the management server 20 can also share information between the control units 11 of the vehicles C by transmitting information received from the control unit 11 of one of the vehicles C to the control unit 11 of the other vehicles C.
• the management server 20 can also transmit information only to a specified control unit 11.
• management system 10 shown in FIG. 6 is illustrated with only one management server 20, there is no restriction on the number of management servers 20 in the management system 10.
• vehicles C having control units 11 may be classified into multiple groups, and a management server 20 may be provided for each group.
• control unit 11> 7 is a diagram showing an example of a functional configuration of the control unit 11 provided in the vehicle C.
• the control unit 11 includes a communication unit 11A, an estimation unit 11B, a control unit 11C, an input unit 11D, and a storage unit 11E.
• the communication unit 11A communicates data with the management server 20 via the Internet 18.
• the estimation unit 11B uses the estimation model M acquired from the management server 20 to estimate the control parameters of the transducer 3 so that the acoustic characteristics inside the vehicle approach the target acoustic characteristics at each point along the time series.
• the control unit 11C controls the vibrator 3 according to the control parameters estimated by the estimation unit 11B.
• the input unit 11D receives measurement values from various sensors included in the sensor group 12, and also receives various information from an information device 19 such as a car navigation device or a camera 17.
• the information device 19 from which the input unit 11D receives various information is not limited to an information device 19 installed in the vehicle C, but may be an information device 19 that an occupant brings into the vehicle when getting in, such as a wearable terminal.
• the input unit 11D stores the received measurement values and information in the memory unit 11E, and notifies at least one of the estimation unit 11B and the control unit 11C, as necessary, that some information has been received from the sensor group 12 and the information device 19.
• the memory unit 11E stores the information received by the input unit 11D, and outputs the information requested by the estimation unit 11B and the control unit 11C to the request source.
• Machine learning is an operation in which a computer learns the causal relationships between data hidden in given sample data, and generates a model that represents the causal relationships between the data. Therefore, the more sample data that is given, the more likely it is that the causal relationships between data hidden in the sample data will be correctly learned.
• the sample data used in machine learning is called "teacher data.”
• FIG. 8 is a diagram showing an example of learning in which machine learning of the estimation model M is performed using teacher data. If there exists an estimation model M that outputs control parameters (also called “desired control parameters") that realize the vibration of the transducer 3 so that the acoustic characteristics of the interior of the vehicle obtained by the vibration of the glass plate structure 9 approach the target acoustic characteristics for input data representing the state inside the vehicle, desired control parameters can be obtained for various states inside the vehicle. However, since there are multiple states inside the vehicle to be considered and they are intricately related to each other, it is often difficult to physically formulate the relationship between the state inside the vehicle and the desired control parameters.
• control parameters also called “desired control parameters”
• teacher data in which the desired control parameters are associated with the input data representing the state inside the vehicle are generated in advance, and when the input data of each teacher data is input, a model that represents the causal relationship between the input data and the output is generated by machine learning only from the relationship between the input data and the output so that the output approaches the control parameter of the corresponding teacher data.
• the method of learning the estimation model M that represents the relationship between the input data and the output using teacher data in which the desired output (in this case, the control parameter) is associated with the input data in this way is called "supervised machine learning”.
• the learning unit 20A of the management server 20 uses multiple pieces of pre-prepared training data to perform supervised machine learning of the estimation model M.
• Deep learning is an example of a neural network that has multiple intermediate layers (also called “hidden layers”), and for example, a convolutional neural network (CNN) is used.
• CNN convolutional neural network
• the input data of the teacher data used in the first embodiment includes, as an example, an attribute representing the space occupancy state inside the vehicle and the opening degree of the glass plate constituent 9 in the front side window FSW and the rear side window RSW.
• the attributes representing the space occupancy state inside the vehicle included in the input data of each teacher data are the position and number of occupants in the vehicle C and the inclination angle ⁇ of the seat 14. Therefore, in the teacher data used for machine learning of the estimation model M, the desired control parameters are associated with the input data including the position and number of occupants in the vehicle C, the inclination angle ⁇ of the seat 14, and the opening degree of the glass plate constituent 9.
• the opening degree of the glass plate constituent 9 refers to the opening degree of each glass plate constituent 9 constituting the front side window FSW and the rear side window RSW.
• the inclination angle ⁇ of the seat 14 refers to the inclination angle ⁇ of each seat 14 provided in the vehicle.
• the distance from the speaker 5 to the occupant and the sound transmission path from the speaker 5 change depending on the position of the occupant in the vehicle C, so the way the sound from the speaker 5 is heard changes when the occupant's position changes.
• the way the sound is heard from the speaker 5 also changes depending on the number of occupants, for example, because the sound is blocked by the occupants and the sound transmission path from the speaker 5 changes.
• the tilt angle ⁇ of the seat 14 changes, the position of the backrest of the seat 14 also changes, so the sound transmission path from the speaker 5 changes and the way the sound from the speaker 5 is heard changes.
• the opening degree of the glass plate structure 9 changes, the area where the target sound is reflected changes, so the way the sound from the speaker 5 is heard changes.
• the position and number of occupants in vehicle C, the inclination angle ⁇ of the seat 14, and the opening degree of the glass plate structure 9 are correlated with the acoustic characteristics inside the vehicle.
• control parameters of the teacher data corresponding to each space occupancy state and the opening degree of the glass plate structure 9 are set in advance by experiments using the vehicle C or computer simulations.
• the control parameters in the teacher data represent the vibration magnitude and vibration phase of the transducer 3 corresponding to a damping sound that dampens the reflected sound from the speaker 5, i.e., the target sound that the occupant wishes to hear, which is reflected at least once inside the vehicle and reaches the occupant, or an emphasis sound that emphasizes the reflected sound.
• control parameters associated with the input data represent the vibration magnitude and vibration phase of the transducer 3 corresponding to a damping sound that dampens the reflected sound
• control parameters associated with the input data may be teacher data that represent the vibration magnitude and vibration phase of the transducer 3 corresponding to an emphasis sound that emphasizes the reflected sound.
• the opening degree for each glass plate structure 9 is included in the input data. Accordingly, teacher data is used that associates the control parameters for each vibrator 3 in each glass vibration plate with vibrator 1 with the input data.
• Fig. 7 can be configured using a computer 30.
• Fig. 10 is a diagram showing an example of the configuration of the main parts of the electrical system of the control unit 11 configured using the computer 30.
• the computer 30 includes a CPU (Central Processing Unit) 31, which is an example of a processor that executes the processing of each functional part in the control unit 11 shown in FIG. 7, a ROM (Read Only Memory) 32 that stores a startup program (Basic Input Output System: BIOS) that performs startup processing of the computer 30, a RAM (Random Access Memory) 33 used as a temporary working area for the CPU 31, a non-volatile memory 34, and an input/output interface (I/O) 35.
• the CPU 31, ROM 32, RAM 33, non-volatile memory 34, and I/O 35 are each connected via a bus 36.
• the non-volatile memory 34 is an example of a storage device that maintains stored information even if the power supplied to the non-volatile memory 34 is cut off.
• a semiconductor memory is used, but a hard disk may also be used.
• a unit 37 having a structure for implementing the functions of the control unit 11 is connected to the I/O 35.
• a communication unit 37A, an input unit 37B, and an output unit 37C are connected to the I/O 35.
• the communication unit 37A is connected to the Internet 18 and has a communication protocol for performing data communication with an external device connected to the Internet 18, such as the management server 20.
• the communication unit 37A may also be connected to an in-vehicle network (e.g., a Controller Area Network (CAN)) in the vehicle C and have a communication protocol for performing data communication with an in-vehicle device connected to the CAN.
• CAN Controller Area Network
• the input unit 37B is an example of a unit 37 that receives the measurement values of each sensor from the sensor group 12 and notifies the CPU 31. Note that the measurement values from a CAN-compatible sensor may also be received from the communication unit 37A.
• the output unit 37C is an example of a unit 37 that outputs information processed by the CPU 31 to the outside of the control unit 11, and produces digital outputs such as contact outputs, and analog outputs such as voltage outputs and current outputs.
• the vibrator 3 of the vibrator-equipped glass diaphragm 1 vibrates according to the control signal output from the output unit 37C.
• the units 37 connected to the I/O 35 shown in FIG. 10 are just an example, and are not necessarily limited to the communication unit 37A, the input unit 37B, and the output unit 37C. Necessary units 37 are connected to the I/O 35 depending on the function of the control unit 11.
• Fig. 11 is a flowchart showing an example of a flow of a control process of the vibrator 3 executed by the CPU 31 of the control unit 11 in accordance with an instruction from an occupant when a sound such as music or radio is being played from the speaker 5 inside the vehicle.
• the information processing program that defines the control process of the vibrator 3 is pre-stored, for example, in the non-volatile memory 34 of the control unit 11.
• the CPU 31 reads the information processing program stored in the non-volatile memory 34 and executes the control process of the vibrator 3.
• the non-volatile memory 34 of the control unit 11 is assumed to have pre-stored therein an estimated model M that has been machine-learned by the learning unit 20A of the management server 20.
• step S10 of FIG. 11 the CPU 31 acquires input information corresponding to each item included in the input data of the teacher data used in the machine learning of the estimation model M from the sensor group 12 and the information device 19.
• the CPU 31 acquires an image of the interior of the vehicle captured by the camera 17, and acquires from the sensor group 12 the opening degree of the glass plate structure 9 measured by a sensor that detects the amount of operation of the power window button, and the inclination angle ⁇ of the seat 14 measured by a sensor that detects the amount of operation of the power seat button that adjusts the angle of the backrest of the seat 14.
• step S12 the CPU 31 uses a known image recognition method to obtain the positions and number of occupants in the vehicle from the image captured by the camera 17. The CPU 31 then inputs the positions and number of occupants in the vehicle C, the inclination angle ⁇ of the seat 14, and the opening degree of the glass plate structure 9 into the estimation model M to estimate the current control parameters for each transducer 3.
• the position and number of occupants in the vehicle can also be obtained using a weight sensor attached to the seat surface of the seat 14.
• step S14 the CPU 31 generates a control signal for each transducer 3 according to the control parameters estimated in step S12, and outputs the generated control signal to the corresponding transducer 3.
• step S16 the CPU 31 determines whether an end command has been received from the occupant. If an end command has not been received, the process proceeds to step S10, and steps S10 to S16 are repeatedly executed until it is determined in the determination process of step S16 that an end command has been received.
• the control unit 11 continuously executes the process of acquiring input information from the sensor group 12 and the information device 19, and outputting a control signal to the vibrator 3 according to the control parameters at the time the input information was acquired.
• control parameters generated by the estimation model M represent the vibration magnitude and vibration phase of the transducer 3 corresponding to the attenuated sound that attenuates the reflected sound of the target sound that the occupant wishes to hear in the state inside the vehicle represented by the input information. Therefore, during the control process of the transducer 3, the reflected sound is attenuated, making the target sound easier to hear compared to when the glass plate structure 9 is not vibrated by the transducer 3.
• control parameters of the teacher data used to generate the estimation model M represent the vibration amount and vibration phase of the transducer 3 corresponding to the emphasis sound that emphasizes the reflected sound of the target sound
• the control parameters generated by the estimation model M will represent the vibration amount and vibration phase of the transducer 3 corresponding to the emphasis sound that emphasizes the reflected sound of the target sound in the state inside the vehicle represented by the input information. Therefore, during the control process of the transducer 3, the reflected sound is emphasized, making the target sound easier to hear compared to when the glass plate structure 9 is not vibrated by the transducer 3.
• step S16 determines whether an end instruction has been received. If it is determined in the determination process of step S16 that an end instruction has been received, the control process of the vibrator 3 shown in FIG. 11 is terminated.
• the positions and number of occupants in the vehicle may be considered not to change until the door of vehicle C is opened again. Therefore, during the control process of vibrator 3, it is not necessary to execute the process of step S10 every time to obtain the positions and number of occupants in the vehicle from the image captured by camera 17; for example, the positions and number of occupants in the vehicle may be obtained from the image captured by camera 17 only the first time.
• some glass vibrating plates 1 with vibrators have multiple vibrators 3 attached to the same glass plate structure 9.
• the CPU 31 generates a control signal for each vibrator 3 attached to the same glass plate structure 9, and outputs the generated control signal to the corresponding vibrator 3 to control the vibration amount and vibration phase of each vibrator 3.
• the vibration of a glass plate structure 9 to which multiple vibrators 3 are attached will have characteristics that are the result of the superposition of vibrations from each vibrator 3. Therefore, for example, it is possible to cooperate in such a way that vibrations in a frequency band that are difficult to generate with one vibrator 3 can be generated by the other vibrator 3.
• the material of the seat 14 is changed, the direction and amount of sound reflected by the seat 14 will change, and the acoustic characteristics inside the vehicle will also change. In other words, the material of the seat 14 is correlated with the acoustic characteristics inside the vehicle.
• the learning unit 20A of the management server 20 may therefore perform machine learning of the estimation model M using teacher data in which the desired control parameters are associated with the input data including the material of the seat 14.
• the CPU 31 acquires the pre-stored material of the seat 14 from the non-volatile memory 34, and in step S12 of FIG. 11, input information including the acquired material of the seat 14 into the estimation model M to estimate the control parameters for each transducer 3.
• the interior surface of the vehicle body that is visible from inside the vehicle is made up of various types of materials, such as plastic materials and fabrics. Furthermore, the types of materials used and where on the interior surface of the vehicle body vary depending on the vehicle model, and may differ depending on the grade even for the same vehicle model.
• the position and type of material that makes up the interior surface of the car body changes, the direction and amount of sound reflected by the interior surface of the car body will change, and the acoustic characteristics inside the car will also change.
• the position and type of material that makes up the interior surface of the car body are correlated with the acoustic characteristics inside the car.
• the learning unit 20A of the management server 20 may therefore perform machine learning of the estimation model M using teacher data in which desired control parameters are associated with input data, including the position and type of material that constitutes the inner surface of the vehicle body.
• the CPU 31 acquires pre-stored data representing the position and type of material that constitutes the inner surface of the vehicle body from the non-volatile memory 34, and in step S12 of FIG. 11, input information including the acquired position and type of material into the estimation model M to estimate the control parameters for each transducer 3.
• the items added to the input data and input information may be either the position or the type of material that constitutes the inner surface of the vehicle body.
• the model of the vehicle C is an example of a model of a moving body.
• the attributes representing the opening degree of the glass plate structure 9 and the state of space occupancy inside the vehicle are the same, there is, for example, music that emphasizes bass and music in which the vocals are difficult to hear. Therefore, if the music played from the speaker 5 is different, the way it is heard may change. In other words, the music played from the speaker 5 is affected by the acoustic characteristics inside the vehicle, and the nature of the effect differs for each type of music.
• the learning unit 20A of the management server 20 may therefore perform machine learning of the estimation model M using teacher data in which desired control parameters set so that the music sounds the way the user prefers in the car are associated with input data plus a music ID that identifies each piece of music.
• the CPU 31 obtains the music ID of the music being played from the speaker 5 from the information device 19 that is playing the music, and in step S12 of FIG. 11, input information including the obtained music ID into the estimation model M to estimate the control parameters for each transducer 3.
• the temperature inside a car changes, so does the amount of moisture that the air can contain as water vapor. For example, the higher the temperature, the more moisture the air can contain per unit volume. As the amount of moisture in the air increases, there is also more material to vibrate, which increases the speed at which sound propagates and makes it easier for sound to travel. In other words, the temperature inside a car correlates with the acoustic characteristics inside the car.
• the humidity inside the car changes, the rate at which sound is absorbed by the moisture in the air changes for each sound frequency band, and the way sound is transmitted changes for each sound frequency band.
• the humidity inside the car is correlated with the acoustic characteristics inside the car.
• the elastic modulus and loss factor of the glass plate structure 9 change with a change in temperature of the glass plate structure 9, and the behavior of the vibration generated in the glass plate structure 9 may change even if the vibration magnitude and phase are the same.
• the learning unit 20A of the management server 20 may perform machine learning of the estimation model M using teacher data in which the desired control parameters are associated with input data including the temperature inside the vehicle, the temperature of the glass plate construct 9, and humidity.
• the CPU 31 acquires the temperature inside the vehicle, the temperature of the glass plate construct 9, and humidity from the sensor group 12, and in step S12 of FIG. 11, input information including the acquired temperature inside the vehicle, the temperature of the glass plate construct 9, and humidity into the estimation model M to estimate the control parameters for each transducer 3.
• the items added to the input data and the input information may be at least one of the temperature inside the vehicle, the temperature of the glass plate construct 9, and humidity.
• the temperature inside the vehicle and the temperature of the glass plate construct 9 may be measured using, for example, a surface thermometer and a thermocouple thermometer.
• the temperature of the glass plate construct 9 may be the temperature of the glass plate construct 9 itself, but the temperature of a member attached to the glass plate construct 9, such as the mount portion 7, may also be used as the temperature of the glass plate construct 9.
• the acoustic characteristics inside the vehicle since sound is heard by the ears, it is preferable for the acoustic characteristics inside the vehicle to be close to the target acoustic characteristics at the head position.
• the learning unit 20A of the management server 20 may therefore perform machine learning of the estimation model M using teacher data in which desired control parameters are associated with input data including head position information of each occupant.
• the CPU 31 acquires head position information of each occupant from an image captured by the camera 17, and in step S12 of FIG. 11, input information including the acquired head position information into the estimation model M to estimate the control parameters for each transducer 3.
• the distance from the vibrator-equipped glass vibration plate 1 that is closest to the occupant to the head is used.
• the front side window FSW is the vibrator-equipped glass vibration plate 1 that is closest to the occupant. Therefore, the distance from the front side window FSW to the position of the occupant's head along the normal direction of the front side window FSW is used as the position information of the occupant's head.
• the front side windows FSW are located on both the left and right sides of the vehicle C, the shortest distance is used as the position information of the occupant's head.
• input data that includes ear position information instead of head position information may be used for the machine learning of the estimation model M.
• the learning unit 20A of the management server 20 may therefore perform machine learning of the estimation model M using teacher data in which the desired control parameters are associated with input data including the operating status of the air conditioner.
• the CPU 31 acquires the operating status of the air conditioner from the sensor group 12, and in step S12 of FIG. 11, input information including the acquired operating status of the air conditioner into the estimation model M to estimate the control parameters for each transducer 3.
• the control unit 11 may not directly use the control parameters obtained from the estimation model M for controlling the transducer 3, but may correct the control parameters depending on the situation and use the corrected control parameters to control the transducer 3. An example of the correction of the control parameters will be described below.
• Each occupant has sound preferences, i.e., preferred acoustic characteristics. For example, occupant A may prefer acoustic characteristics that emphasize bass, while occupant B may prefer acoustic characteristics in which vocals can be clearly heard. Therefore, it is preferable for the control unit 11 to control the transducer 3 so that the acoustic characteristics inside the vehicle approach the preferred acoustic characteristics for each occupant.
• FIG. 14 is a flowchart showing an example of the flow of the initial processing executed by the CPU 31 of the control unit 11.
• the initial processing is executed once, for example, after an occupant gets into the vehicle C. Specifically, the initial processing is executed once after starting up the drive device of the vehicle C, such as the engine or motor.
• the CPU 31 reads the information processing program stored in the non-volatile memory 34 and executes initial processing.
• the non-volatile memory 34 is assumed to have pre-stored therein the facial image and preferred acoustic characteristics of each occupant.
• the preferred acoustic characteristics stored in the non-volatile memory 34 are defined, for example, by the sound pressure for each frequency band. When the sound pressure is increased, the sound in the frequency band corresponding to the increased sound pressure is emphasized, and when the sound pressure is decreased, the sound in the frequency band corresponding to the decreased sound pressure is attenuated, thereby making it possible to set the preferred acoustic characteristics.
• step S20 of FIG. 14 the CPU 31 acquires an image of the interior of the vehicle captured by the camera 17.
• step S22 the CPU 31 uses a known image recognition method to identify the occupants from the image captured by the camera 17. Specifically, it identifies who is in which position by comparing the facial images of the occupants registered in advance in the non-volatile memory 34 with the facial images extracted from the image captured by the camera 17.
• step S24 the CPU 31 acquires the occupant's preferred acoustic characteristics identified by the processing in step S22 from the non-volatile memory 34, and ends the initial processing shown in FIG. 14.
• Initial processing identifies where each passenger is located in the vehicle and obtains each passenger's sound preferences.
• FIG. 15 is a flowchart showing an example of the flow of control processing for the vibrator 3 executed by the CPU 31 of the control unit 11 in accordance with the passenger's instructions when music, radio, or other sounds are being played from the speakers 5 inside the vehicle.
• the CPU 31 reads the information processing program stored in the non-volatile memory 34 and executes the control process for the vibrator 3.
• step S13A The control process for the vibrator 3 shown in FIG. 15 differs from the control process for the vibrator 3 shown in FIG. 11 in that step S13A has been added, and the other processes are the same as the control process for the vibrator 3 shown in FIG. 11. Therefore, the process of step S13A will be described here.
• step S13A After the control parameters of the vibrator 3 corresponding to the input information are estimated by the process of step S12 in FIG. 15, the process proceeds to step S13A.
• step S13A the CPU 31 corrects the control parameters so that the sound from the speaker 5 approaches the preferred acoustic characteristics of each occupant at the position of each occupant identified by the initial processing shown in FIG. 14.
• step S14 the CPU 31 generates a control signal for each transducer 3 according to the control parameters corrected by the processing of step S13A, and outputs the generated control signal to the corresponding transducer 3. Therefore, even if multiple passengers are riding in the same vehicle, each passenger can hear sounds adjusted to their preferred acoustic characteristics.
• the CPU 31 identifies and positions the occupant from the image captured by the camera 17, but the image captured by the camera 17 may be transmitted to the management server 20, and the management server 20 may identify and position the occupant.
• facial images of each occupant are stored in advance in the storage device of the management server 20.
• control parameters estimated by the estimation model M from the input information are corrected to obtain the acoustic characteristics preferred by the occupant, but the learning unit 20A of the management server 20 may perform machine learning of the estimation model M using teacher data in which the desired control parameters are associated with the input data to which the acoustic characteristics preferred by the occupant have been added.
• the control parameters estimated in step S12 of FIG. 15 are already control parameters that reflect the acoustic characteristics preferred by the occupant, so the process of step S13A of FIG. 15 is not necessary.
• an estimation model M that estimates the desired control parameters corresponding to each spatial occupancy state inside the vehicle is machine-learned.
• the luggage 13 cannot be recognized if it is placed in the blind spot of camera 17.
• the acoustic characteristics corresponding to the actual spatial occupancy state inside the vehicle may differ from the acoustic characteristics inside the vehicle that are the premise for estimating the control parameters. Therefore, it is preferable for the control unit 11 to correct the control parameters of transducer 3 estimated using estimation model M to match the actual spatial occupancy state inside the vehicle.
• FIG. 16 is a flowchart showing an example of the flow of an estimation process executed by the CPU 31 of the control unit 11 to estimate the correction amount of a control parameter.
• the estimation process is executed once, for example, after an occupant gets into the vehicle C. Specifically, the estimation process is executed once after the drive device of the vehicle C is started.
• the CPU 31 reads the information processing program stored in the non-volatile memory 34 and executes the estimation process.
• the CPU 31 controls the speaker 5 and the glass diaphragm with vibrator 1 to output the test sound previously stored in the non-volatile memory 34.
• the test sound is a sound source specialized for detecting the occupancy state of the space inside the vehicle, and a sound source is selected that is easier to identify the occupancy state of the space inside the vehicle by its sound than other sound sources.
• the test sound may be a specific music, white noise, pink noise, virtual road noise, virtual wind noise, virtual engine noise, virtual environmental noise, high frequency sound outside the audible range, or a sound source having a specific waveform such as an impulse signal or step signal as well as a sine wave of a specific frequency.
• step S32 the CPU 31 collects the test sound output from the speaker 5 using the microphone 16 and performs a frequency analysis of the collected test sound, for example by performing a short-time Fourier transform on the test sound.
• the frequency distribution of the test sound obtained by the frequency analysis of the test sound is an example of audio information obtained from the test sound.
• the CPU 31 compares the frequency distribution of the test sound with the reference frequency distribution of the test sound, which is the frequency distribution of the test sound collected by the microphone 16 when the same test sound is output from the speaker 5 under the spatial occupancy state that is the premise for generating the estimation model M.
• the reference frequency distribution of the test sound is an example of reference sound information, and is stored in advance in the non-volatile memory 34, for example.
• the difference between the frequency distribution of the test sound obtained by frequency analysis and the pre-stored reference frequency distribution of the test sound represents the difference between the actual spatial occupancy state inside the vehicle and the reference state, which is the spatial occupancy state inside the vehicle assumed when performing machine learning of the estimation model M. Note that while the test sound is being collected, no sounds other than the test sound are output.
• step S34 the CPU 31 estimates the correction amount of the control parameter from the difference between the frequency distribution of the test sound and the reference frequency distribution of the test sound. Specifically, the CPU 31 estimates the correction amount of the control parameter by referring to a correction table in which, for example, the difference between the frequency distribution of the test sound and the reference frequency distribution of the test sound is previously associated with the correction amount of the control parameter output from the estimation model M corresponding to the reference state. This ends the estimation process shown in FIG. 16.
• the frequency distribution of the test sound represents the spatial occupancy state inside the vehicle, so the spatial occupancy state inside the vehicle can be estimated with greater accuracy than when the spatial occupancy state inside the vehicle is estimated from an image of the vehicle interior captured by the camera 17.
• step S13A of the control processing of the transducer 3 shown in FIG. 15 instead of correcting the control parameters so as to approach the acoustic characteristics preferred by each passenger, if the control parameters are corrected using the correction amount of the control parameters obtained by the estimation processing shown in FIG. 16, the target sound becomes easier to hear compared to before the control parameters are corrected.
• Noises that flow into the vehicle interior include, for example, the sound of the drive unit of the vehicle C, road noise generated when the tires touch the ground, wind noise caused by the vehicle C moving, and drumming noise transmitted from the tires to the vehicle body due to unevenness in the road surface. These noises deteriorate the quietness inside the vehicle, so it is desirable to prevent them from flowing into the vehicle interior as much as possible.
• control unit 11 that reduces noise using the glass diaphragm 1 with a vibrator.
• the technique of reducing noise to make the target sound easier to hear is also called “noise canceling.”
• sounds that flow into the vehicle interior are called “ambient sounds.” Noises generated by the movement of the vehicle C, such as the road noise described above, are one example of ambient sounds.
• FIG. 17 is a flowchart showing an example of the flow of noise canceling processing executed by the CPU 31 of the control unit 11 when an instruction to execute noise canceling is received from an occupant.
• the CPU 31 reads the information processing program stored in the non-volatile memory 34 and executes the noise canceling process.
• the noise canceling process shown in FIG. 17 differs from the control process for the vibrator 3 shown in FIG. 11 in that steps S13B to S13D have been added, and the other steps are the same as the control process for the vibrator 3 shown in FIG. 11. Therefore, steps S13B to S13D will be described here.
• step S13B After the control parameters of the vibrator 3 corresponding to the input information are estimated by the process of step S12 in FIG. 17, the process proceeds to step S13B.
• step S13B the CPU 31 collects environmental sound from the microphone 16.
• environmental sound can also be represented by information other than sound.
• a correlation is found between the magnitude of the vibration of the vehicle body and the magnitude of the vibration sound. Since the vibration sound of the vehicle body is also an example of environmental sound, vibration information corresponding to the environmental sound can also be obtained by an acceleration sensor that measures vibration, and the obtained vibration information is an example of environmental sound.
• step S13C the CPU 31 generates an anti-phase signal of the environmental sound collected by the processing of step S13B.
• An anti-phase signal of the environmental sound is a signal that has the same sound pressure as the environmental sound and is in anti-phase with the phase of the environmental sound. "The same sound pressure” means that the difference in sound pressure between the sounds being compared falls within a predetermined range in which they can be considered to be the same sound pressure.
• step S13D the CPU 31 corrects the control parameters so that the antiphase signal of the environmental sound generated by the processing of step S13C is superimposed on the control signal of the vibrator 3 represented by the control parameters estimated by the processing of step S12.
• step S14 the CPU 31 generates a control signal for each transducer 3 according to the control parameters corrected by the processing of step S13D, and outputs the generated control signal to the corresponding transducer 3. Therefore, the environmental sound is cancelled out and the reflected sound of the target sound is attenuated, making it easier to hear the target sound compared to when the noise canceling processing is not executed.
• the learning unit 20A of the management server 20 may perform machine learning of the estimation model M using teacher data in which desired control parameters that cancel out the environmental sound and attenuate the reflected sound of the target sound are associated with the input data to which information about the environmental sound has been added.
• control parameters estimated using the estimation model M are control parameters that reflect the acoustic characteristics that cancel out the environmental sound and attenuate the reflected sound of the target sound, so that if a control signal for each transducer 3 is generated according to the estimated control parameters and the generated control signal is output to the corresponding transducer 3, acoustic characteristics that cancel out the environmental sound and attenuate the reflected sound of the target sound can be obtained without correcting the control parameters.
• the environmental sounds that are the subject of noise cancellation are not limited to those generated by the movement of vehicle C.
• the voices of passersby walking on the sidewalk, the sounds of surrounding vehicles C, and advertising sounds in shopping districts are also examples of environmental sounds that are subject to noise cancellation.
• the target sound may be output from the glass plate component 9 of the glass diaphragm with vibrator 1 without being output from the speaker 5.
• the CPU 31 controls the vibrator 3 so that the target sound, an attenuation sound that attenuates the reflected sound of the target sound, and a cancellation sound of the environmental sound are simultaneously output from the glass plate component 9, the target sound is output from the glass diaphragm with vibrator 1 under conditions in which the reflected sound and the environmental sound are reduced.
• FIG. 18 is a diagram showing an example of a situation in which a target sound is output from the glass plate component 9.
• the example in FIG. 18 shows a situation in which a target sound is output from the glass plate component 9 constituting the front side window FSW.
• the target sound output from the glass plate component 9 constituting the front side window FSW is reflected by the rear window RW
• the transducer 3 attached to the glass plate component 9 constituting the rear window RW by controlling the transducer 3 attached to the glass plate component 9 constituting the rear window RW, the reflected sound from the rear window RW is reduced, making it easier to hear the target sound output from the glass plate component 9 constituting the front side window FSW.
• control parameters in the training data represent the vibration magnitude and vibration phase of the transducer 3 corresponding to the attenuated sound that attenuates the reflected sound that is reflected at least once inside the vehicle and reaches the occupant when the target sound that the occupant wishes to hear is reflected at least once inside the vehicle.
• a control unit 11 that attenuates the environmental sound entering the vehicle interior at the traveling location of the vehicle C by using an estimation model R that outputs characteristic parameters that represent the characteristics of the environmental sound for a combination of vehicle information that represents the traveling state of the vehicle C and environmental information that is information that represents the type of environment in which the vehicle C is traveling.
• vehicle information that represents the traveling state of the vehicle C is an example of mobile body information that represents the moving state of a mobile body.
• the schematic configuration of the vehicle C according to the second embodiment, the functional configuration of the control unit 11, and the main configuration of the electrical system of the control unit 11 are the same as those shown in Figures 5, 7, and 10, respectively, and therefore will not be described.
• the system configuration including the vehicle C equipped with the control unit 11 according to the second embodiment is also the same as that of the management system 10 according to the first embodiment shown in Figure 6.
• the glass vibration plate 1 with a vibrator is applied to the front side window FSW and rear side window RSW of the vehicle C.
• the estimation model R is a model that generates teacher data in advance that associates feature parameters with input data including vehicle information and environmental information, and machine-learns the causal relationship between input data and output solely from the relationship between the input data and output so that when input data for each teacher data is input, the output approaches the feature parameters of the corresponding teacher data. It is generated by the learning unit 20A of the management server 20.
• the vehicle information included in the input data of each teacher data is the speed of vehicle C and the position information of vehicle C.
• the speed of vehicle C can be acquired from a vehicle speed sensor included in the sensor group 12, and the position information of vehicle C can be acquired from, for example, a car navigation device, which is an example of an information device 19.
• the position information of vehicle C is represented by absolute coordinates using, for example, latitude and longitude, but may also be represented by relative coordinates from a reference point whose latitude and longitude are known in advance.
• the position information of vehicle C will be described as being acquired from a car navigation device, but it may also be acquired from a device equipped with a function for measuring position information, such as a smartphone.
• the control unit 11 may be equipped with a GPS sensor and acquire the position information of vehicle C from the GPS.
• meteorological information includes items such as weather, precipitation, temperature, wind speed, and snowfall, and at least one of these items will be used as environmental information.
• the input data of each teaching data set will include precipitation.
• the input data for the training data includes the speed of vehicle C, the position information of vehicle C, and the amount of precipitation.
• the vibration amount of the glass plate structure 9 may be represented by a frequency distribution of the vibration amount indicating the magnitude of vibration for each frequency.
• the characteristic parameters of the training data corresponding to each vehicle information and environmental information are obtained in advance through experiments using vehicle C and computer simulations.
• training data is used that associates the characteristic parameters of the environmental sounds observed by each glass vibration plate with vibrator 1 with the input data.
• Fig. 19 is a flow chart showing an example of the flow of the process of attenuating environmental sound executed by the CPU 31 of the control unit 11 in accordance with an instruction from the occupant.
• the information processing program that specifies the attenuation process of the environmental sound is stored in advance, for example, in the non-volatile memory 34 of the control unit 11.
• the CPU 31 reads the information processing program stored in the non-volatile memory 34 and executes the attenuation process of the environmental sound.
• the non-volatile memory 34 of the control unit 11 stores in advance an estimation model R that has been machine-learned by the learning unit 20A of the management server 20.
• step S40 of FIG. 19 the CPU 31 acquires input information corresponding to each item included in the input data of the teacher data used in the machine learning of the estimation model R from the sensor group 12 and the information device 19.
• the CPU 31 obtains the speed of the vehicle C from a vehicle speed sensor included in the sensor group 12, and obtains the position information of the vehicle C from the car navigation device.
• the CPU 31 also obtains the amount of precipitation from an external device that provides weather information via the Internet 18.
• step S42 the CPU 31 inputs the speed of vehicle C, the position information of vehicle C, and the amount of precipitation acquired in the processing of step S40 into the estimation model R, and estimates the characteristic parameters of the environmental sound expected to occur at the current location represented by the position information of vehicle C.
• step S44 the CPU 31 generates an antiphase signal for the vibration of the glass plate structure 9 represented by the characteristic parameters of the environmental sound estimated by the processing in step S42.
• step S46 the CPU 31 generates a control signal for each transducer 3 according to the antiphase signal generated by the processing in step S44, and outputs the generated control signal to the corresponding transducer 3 to control each transducer 3.
• step S48 the CPU 31 determines whether an end command has been received from the occupant. If an end command has not been received, the process proceeds to step S40, and steps S40 to S48 are repeatedly executed until it is determined in the determination process of step S48 that an end command has been received.
• the control unit 11 acquires input information from the vehicle speed sensor and the car navigation device, and continuously executes a process of controlling the vibrator 3 so as to cancel out the vibration of the glass plate structure 9 represented by the characteristic parameters of the environmental sound expected to be generated at the current location.
• step S48 determines whether an end instruction has been received. If it is determined in the determination process of step S48 that an end instruction has been received, the environmental sound attenuation process shown in FIG. 19 is terminated.
• the environmental sound flowing into the interior of the vehicle C from the glass plate structure 9 is estimated at each position of the vehicle C, and vibrations corresponding to the antiphase signals of the vibrations generated by the environmental sound are generated in the glass plate structure 9. Therefore, the environmental sound flowing into the vehicle interior can be attenuated.
• control unit 11 When the control unit 11 acquires the location information of the vehicle C from a car navigation device, it becomes possible to control the vibrator 3 linked to the map displayed on the car navigation device.
• control unit 11 acquires the location information of the vehicle C from an information device, such as a smartphone or tablet terminal, that the occupant has brought into the car, it becomes possible to control the vibrator 3 linked to the map displayed on the information device.
• some glass vibrating plates 1 with vibrators have multiple vibrators 3 attached to the same glass plate structure 9.
• the CPU 31 generates a control signal for each vibrator 3 attached to the same glass plate structure 9, and outputs the generated control signal to the corresponding vibrator 3, thereby controlling the vibration amount and vibration phase of each vibrator 3.
• the vibration of the glass plate structure 9 to which multiple vibrators 3 are attached will have characteristics that are the superposition of the vibrations caused by each vibrator 3.
• ⁇ Input information used to estimate feature parameters> The above describes an example of estimating feature parameters by inputting the speed of vehicle C, the location information of vehicle C, and the precipitation amount in the area represented by the location information of vehicle C into estimation model R, but the input information used to estimate the feature parameters is not limited to this.
• the vehicle model changes, the shape of the front side window FSW and rear side window RSW of vehicle C will also change, and the tendency of the environmental sound entering the vehicle interior will also change.
• the vehicle model is correlated with the environmental sound entering the vehicle interior.
• the learning unit 20A of the management server 20 may therefore perform machine learning of the estimation model R using teacher data in which the characteristic parameters of the environmental sounds are associated with the input data, including the vehicle model.
• the CPU 31 acquires the vehicle model of the vehicle C that is stored in advance in the non-volatile memory 34, and in step S42 of FIG. 19, input information including the acquired vehicle model into the estimation model R to estimate the characteristic parameters.
• the tendency of environmental sounds entering the car changes depending on the time of day. For example, even if the roads around a park are filled with children's voices during the day, they become quiet at night. In other words, time information is correlated with the environmental sounds entering the car.
• the CPU 31 may estimate the characteristic parameters of the environmental sound using an estimation model R that is generated by including these items in the input data of the training data.
• At least one of the various types of input information used to estimate the control parameters in the first embodiment may be included in the input data of the teacher data to perform machine learning of the estimation model R, and the input information used as input data for the machine learning of the estimation model R may be input to the generated estimation model R to estimate the feature parameters.
• the teacher data of the estimation model R is configured by associating feature parameters with input data including vehicle information and environmental information.
• the traveling vehicle C can acquire vehicle information and environmental information at each traveling point, and can also measure environmental sounds at each traveling point. Therefore, the control unit 11 can generate new teacher data while traveling.
• FIG. 20 is a diagram showing an example of the functional configuration of the control unit 11 corresponding to the generation of teacher data.
• the functional configuration example shown in FIG. 20 differs from the functional configuration example of the control unit 11 shown in FIG. 7 in that a generation unit 11F has been added.
• the generation unit 11F generates teacher data by associating the vehicle information and environmental information acquired at each driving point with the characteristic parameters of the environmental sounds at that driving point. The following describes the operation of the control unit 11 that generates the teacher data.
• FIG. 21 is a flowchart showing an example of the flow of the teacher data generation process executed by the CPU 31 of the control unit 11 when an instruction to generate teacher data is received from the occupant.
• the CPU 31 reads the information processing program stored in the non-volatile memory 34 and executes the process of generating teacher data.
• step S50 the CPU 31 acquires a vibration signal corresponding to the vibration of the glass plate structure 9 from the vibrator 3.
• step S52 the CPU 31 obtains the vibration amount and vibration phase of the glass plate structure 9 from the vibration signal, and generates characteristic parameters of the environmental sound.
• step S54 the CPU 31 acquires input information corresponding to the input data used in the machine learning of the estimation model R. Specifically, if the input data of the estimation model R is the speed of the vehicle C, the position information of the vehicle C, the vehicle model, and the amount of precipitation, the CPU 31 acquires the speed of the vehicle C, the position information of the vehicle C, the vehicle model, and the amount of precipitation at the current location.
• the items of input data used in the machine learning of the estimation model R may be stored in advance in, for example, the non-volatile memory 34.
• step S56 the CPU 31 generates teacher data by associating the input information acquired in the process of step S54 with the characteristic parameters of the environmental sound generated in the process of step S52.
• the CPU 31 stores the generated teacher data in the non-volatile memory 34. Since the input information includes the position information of the vehicle C, the position information is associated with the teacher data.
• step S58 the CPU 31 controls the communication unit 37A to transmit the teacher data generated by the processing of step S56 to the management server 20 via the Internet 18.
• step S60 the CPU 31 determines whether an end instruction has been received from the occupant. If an end instruction has not been received, the process proceeds to step S50, and steps S50 to S60 are repeatedly executed until it is determined in the determination process of step S60 that an end instruction has been received.
• teacher data is generated in which the characteristic parameters of the environmental sounds at the traveling point are associated with the vehicle information and environmental information, and transmitted to the management server 20.
• step S60 if it is determined in the determination process of step S60 that an end instruction has been received, the teacher data generation process shown in FIG. 21 is terminated.
• step S58 and S60 may be reversed, and the determination process of step S60 may be performed before the process of step S58.
• the generated teacher data will be sent to the management server 20 all at once after an end instruction is received from the user, so the load on the CPU 31 will be reduced compared to sending teacher data to the management server 20 every time it is generated.
• the management server 20 which has received the training data from the control unit 11, performs additional machine learning on the estimation model R using the training data received from the control unit 11 in the learning section 20A, thereby updating the estimation model R.
• the management server 20 then transmits the updated estimation model R to the control unit 11 of each vehicle C.
• the CPU 31 of the control unit 11 can estimate the characteristic parameters of the environmental sounds more accurately than when the estimation model R before the update is used.
• training data can be obtained that includes characteristic parameters of environmental sounds corresponding to various vehicle information and environmental information in a wider area than training data generated by a single vehicle C.
• machine learning of the estimation model R is performed by the learning unit 20A of the management server 20, but the control unit 11 may be provided with a local learning unit (not shown) having functions equivalent to the learning unit 20A of the management server 20.
• the learning unit 20A of the management server 20 is unnecessary, and the machine learning of the estimation model R is performed by the local learning unit. Therefore, the CPU 31 can omit the process of transmitting the generated teacher data to the management server 20. It is preferable that the machine learning of the estimation model R by the CPU 31 be performed when the load rate of the CPU 31 is lower than a predetermined threshold, for example, while the vehicle C is stopped.
• the teacher data generated by the teacher data generation process described above includes the position information of vehicle C and characteristic parameters of the environmental sounds generated at the point represented by the position information.
• the generated teacher data it is possible to know the tendency of environmental sounds that are likely to be generated at the driving point of vehicle C.
• FIG. 22 is a flowchart showing an example of the flow of the environmental sound attenuation process executed by the CPU 31 of the control unit 11 when an instruction to attenuate the environmental sound is received from an occupant.
• the CPU 31 reads the information processing program stored in the non-volatile memory 34 and executes the process of attenuating the environmental sound. Note that the teacher data generated by the control unit 11 is assumed to be stored in the non-volatile memory 34 in advance.
• the environmental sound attenuation process shown in FIG. 22 differs from the environmental sound attenuation process shown in FIG. 19 in that steps S45A and S45B have been added, and the other steps are the same as the environmental sound attenuation process shown in FIG. 19. Therefore, the process of steps S45A and S45B will be described here.
• step S44 in FIG. 22 After the processing of step S44 in FIG. 22 generates an antiphase signal for the vibration of the glass plate structure 9 represented by the characteristic parameters of the environmental sound estimated from the estimation model R, the process proceeds to step S45A.
• step S45A the CPU 31 acquires from the non-volatile memory 34 teacher data whose input data contains the same position information as the position information of the vehicle C acquired in step S40.
• the position information of the vehicle C being the same refers to a state in which the difference in the position information of the vehicle C is within a specified range.
• the specified range can be changed by remote control from the management server 20 or by the occupant, and is set to a value such as 25 m, for example.
• step S45B the CPU 31 acquires the characteristic parameters of the teacher data acquired in the processing of step S45A.
• the CPU 31 corrects the antiphase signal generated in the processing of step S44 from the difference between the characteristic parameters acquired from the teacher data and the characteristic parameters estimated by the estimation model R.
• the CPU 31 obtains the correction amount of the antiphase signal by referring to a correction table in which the difference between the feature parameters acquired from the teacher data and the feature parameters estimated by the estimation model R is previously associated with the correction amount of the antiphase signal, and corrects the antiphase signal generated by the processing of step S44.
• the correction amount of the antiphase signal set in the correction table is a correction amount that is pre-adjusted so that the environmental sound flowing into the vehicle C is attenuated more than by controlling the transducer 3 according to the antiphase signal estimated by the estimation model R, for the difference between the corresponding feature parameters.
• step S46 the CPU 31 generates a control signal for each transducer 3 according to the antiphase signal corrected by the processing of step S45B, and outputs the generated control signal to the corresponding transducer 3 to control each transducer 3.
• step S45A of FIG. 22 the CPU 31 acquires from the non-volatile memory 34 teacher data whose input data includes the same position information as the position information of the vehicle C acquired in step S40.
• the CPU 31 may further acquire the current time and acquire from the non-volatile memory 34 teacher data whose input data is closest to the combination of the acquired position information of the vehicle C and the acquired current time.
• teacher data including input data that is closest to the combination of the acquired position information of vehicle C, the acquired current time, and the acquired meteorological information may be acquired from the non-volatile memory 34.
• control unit 11 can obtain a vibration signal corresponding to the vibration of the glass plate structure 9 through the vibrator 3. Therefore, by tapping or sliding the glass plate structure 9 with a finger, it is possible to notify the control unit 11 of an instruction that is pre-associated with the tapped area, or to notify the control unit 11 of an operation amount corresponding to the amount of sliding.
• the occupant may also instruct the control unit 11 on the amount of operation by the strength of the tap on the glass plate structure 9.
• the control unit 11 controls the volume of the information device 19 so that the volume of the speaker 5 corresponds to the strength of the tap on the glass plate structure 9.
• a car navigation device installed in the vehicle C can be operated by a occupant sitting in the seat 14 at the rear of the vehicle C.
• the occupant may also operate the glass plate construct 9 to register and select information about the occupants (referred to as "occupant information"), such as a facial image of each occupant who gets into the vehicle C. Furthermore, if multiple estimation models M or multiple estimation models R exist in the non-volatile memory 34, the occupant may operate the glass plate construct 9 to call up the estimation model M or estimation model R used during the previous drive. The occupant may also operate the glass plate construct 9 to call up the teacher data generated by the generation unit 11F during the previous drive.
• control unit 11 can execute the various processes shown in the flowcharts in Figures 11, 14 to 17, 19, 21, and 22 not only while the moving object is moving, but also when the moving object is stopped.
• processor refers to a processor in a broad sense, and includes, for example, the CPU 31 and dedicated processors.
• Dedicated processors include, for example, a GPU (Graphics Processing Unit), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), and a programmable logic device.
• processor operations in the above embodiments may not only be performed by a single processor, but may also be performed by multiple processors working together in physically separate locations.
• the storage destination of the information processing program is not limited to the non-volatile memory 34.
• the information processing program of the present disclosure can also be provided in a form recorded on a storage medium readable by the computer 30.
• the information processing program may be provided in a form recorded on an optical disc such as a CD-ROM (Compact Disk Read Only Memory), a DVD-ROM (Digital Versatile Disk Read Only Memory), or a Blu-ray disc.
• the information processing program may also be provided in a form recorded on a portable semiconductor memory such as a USB (Universal Serial Bus) memory or a memory card.
• Non-volatile memory, CD-ROMs, DVD-ROMs, Blu-ray discs, USBs, and memory cards are examples of non-transitory storage media.
• control unit 11 may download an information processing program from an external device connected to the Internet 18, such as a management server 20, and store it in the non-volatile memory 34.
• An information processing device including: an estimation unit that estimates the control parameters at each time point from newly acquired input data using an estimation model that estimates the control parameters from the input data, the estimation model being generated by machine learning using teacher data in which control parameters that realize vibration of the oscillator such that acoustic characteristics of the space obtained by vibration of the glass plate structure approach target acoustic characteristics are associated with input data including a space occupancy state that represents the state of a space separated from the surroundings by a housing including a glass vibration plate with an oscillator attached to a glass plate structure having an adjustable opening degree, and an opening degree of the glass plate structure; and a control unit that controls the oscillator in accordance with the control parameters estimated by the estimation unit.
• this information processing device by controlling the glass diaphragm with a vibrator, it is possible to improve the acoustic characteristics in a space surrounded by a housing including the glass diaphragm with a vibrator.
• the attributes representing the space occupancy state include at least one of attributes related to seats in the space and attributes related to baggage brought into the space, as well as the position and number of people in the space.
• the acoustic characteristics in a space surrounded by a housing including a glass diaphragm with a vibrator can be improved according to at least one of seat-related attributes and baggage-related attributes, and the positions and number of people.
• An information processing device in which, when attributes relating to seats in the space are included as attributes representing the space occupancy state, the attributes relating to seats in the space include at least one of seat shape, seat position, and seat material.
• the acoustic characteristics in the space surrounded by the housing including the glass diaphragm with a vibrator can be improved depending on at least one of the shape of the seat, the position of the seat, and the material of the seat.
• the information processing device in which, when attributes related to the luggage are included as attributes representing the spatial occupancy state, the attributes related to the luggage include at least one of a position of the luggage, a shape of the luggage, and a material of the luggage.
• the acoustic characteristics in a space surrounded by a housing including a glass diaphragm with a vibrator can be improved depending on at least one of the position of the luggage, the shape of the luggage, and the material of the luggage.
• the teacher data is a sound emitted from at least one speaker and the glass diaphragm with a vibrator provided in the space
• the teacher data is data in which the vibration amount and the vibration phase of the vibrator corresponding to an attenuated sound that attenuates a reflected sound generated when a target sound that a person in the space wishes to hear is reflected by the housing more than when the glass plate structure is not vibrated, or an emphasized sound that emphasizes the reflected sound of the target sound more than when the glass plate structure is not vibrated, are associated with the input data as the control parameters
• the control unit controls the vibration amount and the vibration phase of the vibrator for each of at least one vibrator attached to the glass plate structure in accordance with the control parameters estimated from the estimation model generated using the teacher data.
• the sound reflected by the housing can be attenuated more effectively than in a case where the glass plate structure is not vibrated by the vibrator.
• control parameters are a vibration magnitude and a vibration phase of the vibrator corresponding to the attenuating sound that attenuates a reflected sound of the target sound that has been reflected at least once by the housing, or the emphasizing sound that emphasizes a reflected sound of the target sound that has been reflected at least once by the housing.
• control unit identifies people in the space from an image taken of the space, obtains the sound preferences of the identified people from preference information for each person previously stored in a storage device, corrects the control parameters so that the sound from the speaker at the position of each identified person approaches the sound preferences of each identified person, and controls the vibration amount and vibration phase of the vibrator in accordance with the corrected control parameters.
• this information processing device it is possible to realize acoustic characteristics that suit the preferences of each person.
• control unit acquires environmental sound collected by a microphone, generates an anti-phase signal having the same sound pressure as the environmental sound and an anti-phase signal having an opposite phase to the phase of the environmental sound, and corrects the control parameter so that the anti-phase signal is superimposed on a control signal for the vibrator based on a control parameter estimated from the estimation model.
• the control unit acquires as the environmental sound at least one of the sound of the vehicle's drive source, road noise generated by the tires contacting the ground, wind noise caused by the moving body, and drumming noise transmitted from the tires to the housing due to unevenness in the road surface. According to this information processing device, it is possible to reduce noise entering the interior of the vehicle while the vehicle is running.
• test sound is music predetermined for use in estimating the spatial occupancy state, or a sound represented by an impulse signal. According to this information processing device, it becomes easier to set the correction amount of the control parameter compared to using an arbitrary sound as the test sound.
• the estimation unit estimates the control parameter at each point in time from newly acquired input data using the estimation model generated by machine learning to estimate the control parameter from the input data including the type of material constituting the inner surface of the housing.
• the acoustic characteristics in the space surrounded by the housing including the glass diaphragm with a vibrator can be improved depending on the type of material that constitutes the inner surface of the housing.
• the input data further includes at least one of a temperature of the space, a temperature of the glass plate structure, and humidity
• the estimation unit estimates the control parameter at each time point from newly acquired input data using the estimation model generated by machine learning to estimate the control parameter from the input data including at least one of the temperature of the space, the temperature of the glass plate structure, and humidity.
• the acoustic characteristics in a space enclosed by a housing including a glass diaphragm with a vibrator can be improved in response to at least one of the temperature and humidity of the space.
• An information processing device comprising: an estimation unit that estimates feature parameters at each point from newly acquired input data using an estimation model that estimates feature parameters from the input data, the feature parameters being generated by machine learning using teacher data in which input data includes vehicle information that represents a driving state of a vehicle at a point where the vehicle is located, the vehicle having a glass vibration plate with a vibrator attached to a glass plate structure and environmental information that represents the environment at the point where the vehicle is located, and feature parameters that represent characteristics of environmental sounds flowing into the interior of the vehicle from outside the vehicle at the point where the vehicle is located; and a control unit that controls the vibrator so as to attenuate the environmental sounds flowing into the interior of the vehicle, which are represented by the feature parameters estimated by the estimation unit.
• this information processing device by controlling the glass diaphragm with a vibrator, it is possible to improve the acoustic characteristics in a space surrounded by a housing including the glass diaphragm with a vibrator.
• the information processing device according to (24), wherein the vehicle information of the vehicle includes a speed of the vehicle and position information of the vehicle. According to this information processing device, it is possible to improve the acoustic characteristics in a vehicle interior surrounded by a vehicle body equipped with a glass diaphragm with a vibrator, in accordance with the vehicle speed and vehicle position information.
• the information processing device according to (26), wherein the meteorological information includes at least one of weather, precipitation, temperature, wind speed, and snowfall.
• the acoustic characteristics in a vehicle interior surrounded by a vehicle body equipped with a glass diaphragm with a vibrator can be improved in accordance with at least one of the weather, the amount of precipitation, the wind speed, and the amount of snowfall.
• the characteristic parameters are a vibration magnitude and a vibration phase of the glass plate structure generated by an environmental sound flowing from outside the vehicle into the inside of the vehicle through the glass plate structure
• the estimation unit estimates a vibration magnitude and a vibration phase of the glass plate structure at a point where the vehicle is located from the estimation model generated using the teacher data in which the vibration magnitude and the vibration phase of the glass plate structure are associated with the input data
• the control unit generates an anti-phase signal having the same vibration magnitude as the estimated vibration magnitude of the glass plate structure and an anti-phase to the estimated vibration phase of the glass plate structure, and controls the vibrator so that the glass plate structure vibrates in accordance with the generated anti-phase signal.
• this information processing device it is possible to attenuate environmental sounds flowing from outside the vehicle into the inside of the vehicle, compared to a case in which the vibrator attached to the glass plate structure is not controlled.
• the information processing device according to any one of (28) to (30), further comprising a generation unit that generates the teacher data by associating the vibration magnitude and vibration phase of the glass plate structure acquired through the vibrator with the input data obtained at the point where the vibration magnitude and vibration phase of the glass plate structure were acquired, and the control unit controls output of the teacher data generated by the generation unit to a learning unit that performs machine learning on the estimation model.
• new teacher data can be generated as the vehicle travels.
• the information processing device according to (31) or (32), further comprising the learning unit, wherein the learning unit performs additional learning of the estimation model by using the teacher data generated by the generation unit, and updates the estimation model.
• the estimation model can be updated even when the information processing device is not connected to the Internet.
• control unit outputs the teacher data generated by the generation unit to an external device including the learning unit, and acquires from the external device the updated estimation model in which additional learning has been performed using the teacher data generated by the generation unit, and the estimation unit estimates the feature parameters at each point from the newly acquired input data using the updated estimation model.
• the hardware performance required for the information processing device can be kept lower than the performance required for performing machine learning of the estimation model.
• control unit acquires the teacher data corresponding to a position of the vehicle from the teacher data generated by the generation unit, and performs control to correct an anti-phase signal using a vibration amount and a vibration phase of the glass sheet structure included in the acquired teacher data so that environmental sound flowing into the inside of the vehicle is attenuated more than by controlling the vibrator according to an anti-phase signal generated from the vibration amount and the vibration phase of the glass sheet structure estimated by the estimation unit.
• this information processing device it is possible to attenuate environmental noise at the vehicle's traveling location, compared to a case in which the inverse phase signal is corrected using the vibration magnitude and vibration phase of a glass plate structure that is not associated with position information.
• the estimation unit estimates the feature parameters at each point from the newly acquired input data using the estimation model generated by machine learning to estimate the feature parameters from the input data including the opening degree of the glass plate structure.
• the acoustic characteristics in the interior of a vehicle surrounded by a vehicle body equipped with a glass diaphragm with a vibrator can be improved depending on the degree of opening of the glass plate structure.
• the input data further includes a spatial occupancy state representing an interior state of the vehicle
• the estimation unit estimates the feature parameters at each point from the newly acquired input data using the estimation model generated by machine learning to estimate the feature parameters from the input data including the spatial occupancy state
• the attributes representing the spatial occupancy state include at least one of attributes related to seats inside the moving body and attributes related to baggage brought into the moving body, as well as the positions and number of people riding inside the moving body.
• the acoustic characteristics in the vehicle interior surrounded by the vehicle body equipped with the glass diaphragm with a vibrator can be improved according to at least one of the attributes related to the seats and the attributes related to the luggage, and the positions and the number of people riding inside the vehicle.
• the information processing device according to (38), wherein when attributes related to seats inside the vehicle are included as attributes representing the spatial occupancy state, the attributes related to seats inside the vehicle include at least one of a seat shape and a seat material.
• the acoustic characteristics in a vehicle interior surrounded by a vehicle body equipped with a glass diaphragm with a vibrator can be improved depending on at least one of the shape of the seat and the material of the seat.
• the information processing device in which when attributes related to the luggage are included as attributes representing the spatial occupancy state, the attributes related to the luggage include at least one of a position of the luggage, a shape of the luggage, and a material of the luggage.
• the acoustic characteristics in a vehicle interior surrounded by a vehicle body equipped with a glass diaphragm with a vibrator can be improved according to at least one of the position of the luggage, the shape of the luggage, and the material of the luggage.
• An information processing program for causing a computer to execute a process of estimating the control parameters at each time point from newly acquired input data using an estimation model that estimates the control parameters from the input data, the estimation model being generated by machine learning using teacher data in which input data including a space occupancy state representing a state of a space separated from the surroundings by a housing including a glass diaphragm with a vibrator in which a vibrator is attached to a glass plate structure having an adjustable opening degree, and an opening degree of the glass plate structure, is associated with control parameters that realize vibration of the vibrator such that acoustic characteristics of the space obtained by vibration of the glass plate structure approach target acoustic characteristics, and the information processing program causes a computer to execute a process of estimating the control parameters at each time point from newly acquired input data and controlling the vibrator in accordance with the estimated control parameters.
• this information processing program by controlling the glass diaphragm with a vibrator, it is possible to improve the acoustic characteristics in a space surrounded by a housing including
• An information processing program for causing a computer to execute a process of estimating the feature parameters at each point from newly acquired input data using an estimation model that estimates the feature parameters from the input data, the feature parameters being generated by machine learning using teacher data in which input data includes input data including vehicle information representing the running state of a vehicle at a point where the vehicle is located, the vehicle body having a glass vibration plate with a vibrator attached to a glass plate structure, and environmental information representing the environment at the point where the vehicle is located, and feature parameters representing the characteristics of environmental sounds flowing into the inside of the vehicle from outside the vehicle at the point where the vehicle is located, and controlling the vibrator so as to attenuate the environmental sounds flowing into the inside of the vehicle represented by the estimated feature parameters.
• this information processing program by controlling the glass diaphragm with a vibrator, it is possible to improve the acoustic characteristics in a space surrounded by a housing including the glass diaphragm with a vibrator.

Landscapes

• Engineering & Computer Science (AREA)
• Physics & Mathematics (AREA)
• Acoustics & Sound (AREA)
• Signal Processing (AREA)
• Multimedia (AREA)
• Mechanical Engineering (AREA)
• Fittings On The Vehicle Exterior For Carrying Loads, And Devices For Holding Or Mounting Articles (AREA)

Priority Applications (1)

Applications Claiming Priority (2)

Publications (1)

Family

ID=92500571

Family Applications (1)

Country Status (2)

Cited By (2)

Citations (4)

• 2024
  • 2024-02-07 WO PCT/JP2024/004085 patent/WO2024176827A1/ja not_active Ceased
  • 2024-02-07 JP JP2025502251A patent/JPWO2024176827A1/ja active Pending

Patent Citations (4)

Also Published As

Similar Documents

Legal Events