Abnormal sound occurrence factor determination method and abnormal sound occurrence factor determination device

Abstract

The present disclosure provides an abnormal sound occurrence factor determination method and an abnormal sound occurrence factor determination apparatus. In a characteristic correction process, a sound signal obtained through sound signal obtaining processing is corrected based on obtained model information, so that a frequency characteristic of the obtained sound signal approaches a frequency characteristic of a learning sound signal. In a variable obtaining process, a variable output from a map is obtained by inputting the corrected sound signal to the map. In a factor determination process, an occurrence factor of a sound perceived by a microphone is determined based on the variable obtained through the variable obtaining process (FIG. 5).

Description

Technical Field

[0001] This disclosure relates to a method and apparatus for determining the causes of abnormal noise. Background Technology

[0002] Japanese Patent Application Publication No. 2021-154816 discloses a mapping obtained by performing machine learning to infer the location of factors that contribute to sound generated in a vehicle. This discloses a technique for determining the location of factors that contribute to sound perceived by a microphone.

[0003] In this method, the actuator inputs a sound signal (which is a signal related to the sound sensed by the microphone) and a state variable of the vehicle's drive system into a mapping, and obtains a variable output from the mapping. Furthermore, the actuator determines the location of the sound-sensing factor based on the variable output from the mapping. Summary of the Invention

[0004] According to one aspect of this disclosure, a first example of a method for determining the causes of abnormal noise is provided. The method includes storing mapping data of a predetermined mapping via a storage circuit of a parsing device. The mapping input is an audio signal as a signal related to a sound perceived by a microphone, and the output is a variable related to the causes of sound in a vehicle. The mapping is obtained by performing machine learning. The audio signal input to the mapping during machine learning is a learning audio signal. The microphone that perceives the sound represented by the learning audio signal is a learning microphone. The method further includes: performing an audio signal acquisition process via an execution circuit of the parsing device, in which the audio signal acquisition process acquires the audio signal related to the sound perceived by the microphone; and acquiring vehicle model information via the execution circuit, the vehicle model information being information related to the model of the microphone. The method further includes performing a characteristic correction process via the execution circuit, in which the characteristic correction process corrects the audio signal acquired by the audio signal acquisition process based on the acquired vehicle model information, thereby making the frequency characteristics of the audio signal closer to the frequency characteristics of the learning audio signal. The method further comprises: performing a variable acquisition process through the execution circuit, wherein the variable acquisition process acquires a variable output from the mapping by inputting the sound signal corrected by the characteristic correction process into the mapping; and performing a factor determination process, wherein the factor of sound generation perceived by the microphone is determined based on the variable acquired through the variable acquisition process.

[0005] In the above-described method for determining the causes of abnormal noise, the frequency characteristics of the sound signal related to the sound perceived by the microphone are corrected according to the microphone model. This reduces the deviation in the frequency characteristics of the sound signal caused by differences in the microphone models that perceive the sound. In other words, the frequency characteristics of the sound signal input to the mapping can be made closer to the frequency characteristics of the learning sound signal. Furthermore, by inputting the corrected sound signal to the mapping, the cause of the sound perceived by the microphone is determined based on the variables output from the mapping. Therefore, the accuracy deviation in determining the cause of sound corresponding to the microphone model can be reduced.

[0006] In this method, a microphone used to acquire a learning sound signal as input to the mapping during machine learning is designated as the learning microphone. The model of the microphone used to sense sounds occurring in the vehicle is sometimes different from the model of the learning microphone. The frequency characteristics of the microphone model are reflected in the sound signal. Therefore, when the model of the microphone used to sense sounds occurring in the vehicle differs from the model of the learning microphone, the frequency characteristics of the sound signal related to the sound sensed by the microphone will deviate from the frequency characteristics of the learning sound signal. Therefore, it is difficult to say that the accuracy of determining the location of sound occurrence based on the variables output from the above mapping is high. The above method reduces this phenomenon.

[0007] According to another aspect of this disclosure, a second example of a method for determining the causes of abnormal noise is provided. The method for determining the causes of abnormal noise includes storing mapping data of a predetermined mapping via a storage circuit of a parsing device. The mapping takes an audio signal, which is a signal related to sound perceived by a microphone, as input, and outputs variables related to the causes of sound in a vehicle. The mapping is obtained by performing machine learning. The audio signal input to the mapping during machine learning is a learning audio signal. The microphone that perceives the sound represented by the learning audio signal is a learning microphone. The method further includes: performing an audio signal acquisition process via an execution circuit of the parsing device, in which the audio signal related to sound perceived by the microphone is acquired; and acquiring vehicle model information via the execution circuit, the vehicle model information being information related to the model of the microphone. The method further includes: performing a first characteristic correction process by the execution circuit to correct the frequency characteristics of the sound signal obtained through the sound signal acquisition process, and when the model information of the microphone is first model information, the first characteristic correction process enables the frequency characteristics of the sound signal to be close to the frequency characteristics of the learning sound signal; and performing a second characteristic correction process by the execution circuit to correct the frequency characteristics of the sound signal obtained through the sound signal acquisition process, and when the model information of the microphone is second model information, the second characteristic correction process enables the frequency characteristics of the sound signal to be close to the frequency characteristics of the learning sound signal. The method further includes performing a variable acquisition process by the execution circuit, in which a variable output from the mapping is obtained as a first output variable by inputting the sound signal corrected by the first characteristic correction process to the mapping, a variable output from the mapping is obtained as a second output variable by inputting the sound signal corrected by the second characteristic correction process to the mapping, and a variable output from the mapping is obtained as a third output variable by inputting the sound signal obtained through the sound signal acquisition process to the mapping. The method further includes performing a factor selection process by the execution circuit, in which the sound generation factor is selected from the sound generation factors based on the first output variable, the sound generation factors based on the second output variable, and the sound generation factors based on the third output variable.

[0008] In the above method for determining the cause of abnormal noise, a first characteristic correction process and a second characteristic correction process are performed. Next, a variable acquisition process is performed to obtain a first output variable, a second output variable, and a third output variable. Furthermore, the cause of the sound is selected from the causes determined based on the first output variable, the second output variable, and the third output variable. For example, compared to obtaining only one of the causes determined based on the first, second, and third output variables and identifying the obtained cause as the cause, the accuracy of determining the cause of the sound acquired by the microphone is less likely to decrease in the above method. Therefore, the deviation in the accuracy of determining the cause of the sound corresponding to the microphone model can be reduced.

[0009] According to another aspect of this disclosure, a first example of an apparatus for determining the causes of abnormal noise is provided. The method for determining the causes of abnormal noise determines the causes of sound perceived by a microphone. The apparatus includes an execution circuit and a storage circuit. The storage circuit stores mapping data of a predetermined mapping. The mapping takes an audio signal, which is a signal related to the sound perceived by the microphone, as input, and outputs a variable related to the causes of sound in a vehicle. The mapping is obtained by performing machine learning. The audio signal input to the mapping during machine learning is a learning audio signal. The microphone that perceives the sound represented by the learning audio signal is a learning microphone. The execution circuit performs a characteristic correction process, in which the frequency characteristics of the audio signal related to the sound perceived by the microphone are made close to the frequency characteristics of the learning audio signal by correction corresponding to model information, which is information related to the model of the microphone. In the variable acquisition process, the variable output from the mapping is acquired by inputting the audio signal corrected by the characteristic correction process into the mapping. In the factor determination process, the factors that cause the sound perceived by the microphone are determined based on the variables obtained through the variable acquisition process.

[0010] The above-mentioned device for determining the causes of abnormal noise can have the same function and effect as the first example of the above-mentioned method for determining the causes of abnormal noise.

[0011] According to another aspect of this disclosure, a second example of an apparatus for determining the causes of abnormal noise is provided. The apparatus determines the causes of sound perceived by a microphone. The apparatus includes an execution circuit and a storage circuit. The storage circuit stores mapping data with a predetermined mapping. The mapping takes an audio signal, which is a signal related to the sound perceived by the microphone, as input, and outputs a variable related to the causes of sound in the vehicle. The mapping is obtained by performing machine learning. The audio signal input to the mapping during machine learning is a learning audio signal. The microphone that perceives the sound represented by the learning audio signal is a learning microphone. The execution circuit performs a first characteristic correction process to correct the frequency characteristics of the audio signal related to the sound perceived by the microphone. The first characteristic correction process is a process that, when the microphone model information is first model information, makes the frequency characteristics of the audio signal close to the frequency characteristics of the learning audio signal. A second characteristic correction process is a process that corrects the frequency characteristics of the audio signal. The second characteristic correction process is a process that, when the microphone model information is second model information, enables the frequency characteristics of the sound signal to approximate the frequency characteristics of the learning sound signal. In the variable acquisition process, by inputting the sound signal corrected by the first characteristic correction process into the mapping, a variable output from the mapping is acquired as a first output variable. In the variable acquisition process, by inputting the sound signal corrected by the second characteristic correction process into the mapping, a variable output from the mapping is acquired as a second output variable. In the variable acquisition process, by inputting the uncorrected sound signal into the mapping, a variable output from the mapping is acquired as a third output variable. In the factor selection process, a sound generation factor is selected from the sound generation factors based on the first output variable, the sound generation factors based on the second output variable, and the sound generation factors based on the third output variable.

[0012] The above-mentioned device for determining the causes of abnormal noise can have the same function and effect as the second example of the above-mentioned method for determining the causes of abnormal noise. Attached Figure Description

[0013] Figure 1 This is a block diagram illustrating the structure of the system according to the first embodiment of this disclosure.

[0014] Figure 2 It is shown Figure 1 The table shows the model data for the microphones.

[0015] exist Figure 3 In the middle, part (A) shows the result of... Figure 1The flowchart shown illustrates a series of processes performed by the vehicle control unit. Figure 3 In the middle, part (B) shows the... Figure 1 The flowchart shown is a series of processes performed by the portable terminal.

[0016] Figure 4 It is shown that... Figure 1 A diagram showing an example of a sound signal related to the sound sensed by the microphone of a portable terminal.

[0017] Figure 5 It is shown by Figure 1 The flowchart shown is a part of a series of processes performed by the central control device.

[0018] Figure 6 then Figure 5 This is a flowchart showing the remainder of a series of processes performed by a central control unit.

[0019] Figure 7 It shows the... Figure 1 The diagram shown is a block diagram of the structure of a learning device that implements machine learning.

[0020] Figure 8 It is a replacement Figure 1 A block diagram showing the structure of the system according to the second embodiment is provided. Detailed Implementation

[0021] It is hoped that the description "at least one of A and B" in this specification means "only A" or "only B" or "both A and B".

[0022] The following is in accordance with Figures 1 to 7 This describes the first embodiment of the method for determining the causes of abnormal noise, the processing for determining the causes of abnormal noise, and the device for determining the causes of abnormal noise.

[0023] Figure 1 The diagram shows a vehicle 10, a portable terminal 30 held by the occupants of the vehicle 10, and a data analysis center 60 located outside the vehicle 10.

[0024] <Vehicles>

[0025] The vehicle 10 is equipped with a detection system 11, a vehicle communication unit 13, and a vehicle control device 15.

[0026] The detection system 11 has N sensors 111, 112, 113, ..., 11N. "N" is an integer greater than or equal to 4. The sensors 111 to 11N output signals corresponding to the detection results to the vehicle control unit 15. The sensors 111 to 11N include sensors that detect vehicle state quantities such as vehicle speed and acceleration, and sensors that detect occupant operation quantities such as accelerator operation quantities and brake operation quantities. Furthermore, the sensors 111 to 11N can include sensors that detect the operating state of the vehicle 10's drive system, such as the engine and electric motor, and can also include sensors that detect the temperature of the coolant and oil.

[0027] The vehicle communication unit 13 communicates with the portable terminal 30 that has been brought into the vehicle 10's interior. The vehicle communication unit 13 outputs information received from the portable terminal 30 to the vehicle control unit 15, or sends information output from the vehicle control unit 15 to the portable terminal 30.

[0028] The vehicle control unit 15 controls the vehicle 10 based on the output signals of multiple sensors 111 to 11N. That is, the vehicle control unit 15 controls the vehicle 10's speed, acceleration, and yaw rate by causing the vehicle 10's drive system, braking system, and steering system to operate.

[0029] The vehicle control unit 15 includes a vehicle CPU 16, a first storage device 17, and a second storage device 18. The first storage device 17 is a storage circuit that stores various control programs executed by the vehicle CPU 16. Additionally, the first storage device 17 also stores vehicle model information, such as information related to the vehicle model and class of the vehicle 10. The second storage device 18 is a storage circuit that stores the calculation results of the vehicle CPU 16.

[0030] Portable terminal

[0031] The portable terminal 30 is, for example, a smartphone or tablet. The portable terminal 30 includes a touch panel 31, a display screen 33, a microphone 35, a terminal communication device 37, and a terminal control device 39. The touch panel 31 is a user interface that overlaps with the display screen 33. When the portable terminal 30 is brought into the vehicle cabin, the microphone 35 can detect sound propagating within the cabin.

[0032] The terminal communication device 37 has the function of communicating with the vehicle 10 when the portable terminal 30 is inside the vehicle 10. The terminal communication device 37 outputs information received from the vehicle control device 15 to the terminal control device 39, or sends information output by the terminal control device 39 to the vehicle control device 15.

[0033] In addition, the terminal communication device 37 has the function of communicating with other portable terminals 30 and data analysis center 60 via the global network 100. The terminal communication device 37 outputs information received from other portable terminals 30 or data analysis center 60 to the terminal control device 39, or sends information output by the terminal control device 39 to other portable terminals 30 or data analysis center 60.

[0034] The terminal control device 39 includes a terminal CPU 41, a first storage device 42, and a second storage device 43. In this embodiment, the terminal control device 39 constitutes an example of a "parse device." Furthermore, the terminal CPU 41 constitutes an example of an "execution circuit of the parsing device." An execution circuit corresponds to an execution device. The terminal CPU 41 corresponds to the "first execution circuit." The first execution circuit corresponds to the first execution device. The first storage device 42 is a storage circuit that stores various control programs executed by the terminal CPU 41. Additionally, the first storage device 42 also stores model information, which relates to the model of the microphone 35 installed in the portable terminal 30. The second storage device 43 is a storage circuit that stores the calculation results, etc., of the terminal CPU 41.

[0035] <Data Analysis Center>

[0036] The data analysis center 60 corresponds to the "abnormal noise occurrence factor determination device" that determines the factors causing the sound sensed by the microphone 35. Assume there are M possible factors causing abnormal noise in the vehicle 10. "M" is an integer greater than or equal to 2. In this case, the data analysis center 60 selects one factor from the M candidate factors.

[0037] The data analysis center 60 is equipped with a central communication unit 61 and a central control device 63.

[0038] The central communication unit 61 has the function of communicating with multiple portable terminals 30 via the global network 100. The central communication unit 61 outputs information received from the portable terminals 30 to the central control device 63, or sends information output by the central control device 63 to the portable terminals 30.

[0039] The central control device 63 includes a central CPU 64, a first storage device 65, and a second storage device 66. In this embodiment, the central control device 63 constitutes an example of a "analysis device". Furthermore, the central CPU 64 constitutes an example of an "execution circuit of the analysis device", and the central CPU 64 corresponds to the "second execution circuit". The second storage device 66 corresponds to the "storage circuit of the analysis device". Moreover, the central CPU 64 corresponds to the "execution circuit of the abnormal noise generation factor determination device", and the second storage device 66 corresponds to the "storage circuit of the abnormal noise generation factor determination device".

[0040] The first storage device 65 is a storage circuit that stores various control programs executed by the central CPU 64.

[0041] The second storage device 66 is a storage circuit that stores mapping data 71, which specifies the mapping obtained through machine learning. The mapping is a learned model that outputs variables used to determine the factors contributing to sound generation in the vehicle 10 when input variables are input into it. An example of a mapping is a function approximator. For instance, the mapping is a fully connected, forward-propagating neural network with one intermediate layer.

[0042] Explain the output variable y of the mapping. As mentioned above, there are M candidate factors for the occurrence of abnormal noise in vehicle 10. Therefore, when the input variable is input into the mapping, M output variables y(1), y(2), ..., y(M) are output from the mapping. When the actual occurrence factor is taken as the actual factor, the output variable y(1) represents the probability that the first occurrence factor among the M candidate factors is the actual factor. The output variable y(2) represents the probability that the second occurrence factor among the M candidate factors is the actual factor. The output variable y(M) represents the probability that the Mth occurrence factor among the M candidate factors is the actual factor.

[0043] The second storage device 66 is a storage circuit that stores factor determination data 72. Factor determination data 72 is used to determine the sound generation factors in vehicle 10 based on the mapped output variable y. M generation factor candidates are stored in factor determination data 72. The first generation factor candidate among the M generation factor candidates corresponds to the output variable y(1). The second generation factor candidate among the M generation factor candidates corresponds to the output variable y(2). The Mth generation factor candidate among the M generation factor candidates corresponds to the output variable y(M).

[0044] The second storage device 66 stores model data 73. Model data 73 includes model information for various types of microphones.

[0045] Figure 2 The figure shows an example of model data 73. Figure 2 The model data 73 shown includes the following microphone model information.

[0046] ■ This indicates that the microphone frequency characteristic of AA Communications' portable terminal model "T778" is "A characteristic".

[0047] ■ This indicates that the microphone frequency characteristic of AA Communications' portable terminal model "T548" is "B characteristic".

[0048] ■ This indicates that the microphone frequency characteristics of the BB Mobile Service portable terminal model "M458" are "A characteristic improvement".

[0049] ■ This indicates that the microphone frequency characteristic of the BB Mobile Service portable terminal model "M241" is "A characteristic".

[0050] ■ This indicates that the microphone frequency characteristics of the CC Communication portable terminal model "D111" are "B characteristic improvement".

[0051] ■ This indicates that the microphone frequency characteristic of the CC Communication portable terminal model "D211" is "A characteristic".

[0052] ■ This indicates model information for other microphone models, such as "Type 23", whose frequency characteristics are "F characteristics".

[0053] Here, the frequency range of sounds that are easily perceived by the microphone and the frequency range of sounds that are difficult for the microphone to perceive vary depending on the microphone model. Such microphone characteristics are equivalent to "microphone frequency characteristics".

[0054] As detailed later, let's assume that the microphone of model "23" is used during machine learning mapping. In this case, the microphone of model "23" corresponds to "Learning Microphone 35A" (see reference). Figure 7 ).

[0055] <Methods for Determining the Factors Causing Abnormal Noises>

[0056] Reference Figures 3-6 This explains the method for determining the causes of abnormal noises. Figure 3 Part (A) illustrates the processing flow executed by the vehicle CPU 16 of the vehicle control unit 15. The control program stored in the first storage device 17 is executed repeatedly by the vehicle CPU 16. Figure 3 The series of processes shown in section (A) is as follows.

[0057] exist Figure 3 In the series of processes shown in section (A), in step S11, the vehicle CPU 16 determines whether synchronization with the portable terminal 30 has been established. If the vehicle CPU 16 determines that synchronization with the portable terminal 30 has been established (S11: "Yes"), the process proceeds to step S13. On the other hand, if the vehicle CPU 16 determines that synchronization with the portable terminal 30 has not been established (S11: "No"), the series of processes is temporarily terminated.

[0058] In step S13, the vehicle CPU 16 determines whether the vehicle model information of the vehicle 10 has been sent to the portable terminal 30. If the vehicle model information of the vehicle 10 has been sent to the portable terminal 30 (S13: "Yes"), the vehicle CPU 16 transfers the processing to step S17. On the other hand, if the vehicle model information of the vehicle 10 has not been sent to the portable terminal 30 (S13: "No"), the vehicle CPU 16 transfers the processing to step S15. In step S15, the vehicle CPU 16 sends the vehicle model information of the vehicle 10 from the vehicle communication unit 13 to the portable terminal 30. Afterwards, the vehicle CPU 16 transfers the processing to step S17.

[0059] In step S17, the vehicle CPU 16 acquires the state variables of the vehicle 10. Specifically, the vehicle CPU 16 acquires the detection values ​​from various sensors 111 to 11N, as well as the processed values ​​obtained from processing the detection values, and uses them as the state variables of the vehicle 10. For example, the vehicle CPU 16 acquires the vehicle 10's driving speed SPD, vehicle 10's acceleration G, engine speed NE, and engine torque Trq as the state variables of the vehicle 10.

[0060] In step S19, the vehicle CPU 16 sends the acquired state variables of the vehicle 10 from the vehicle communicator 13 to the portable terminal 30. Afterwards, the vehicle CPU 16 temporarily terminates a series of processes.

[0061] Figure 3 Part (B) illustrates the processing flow executed by the terminal CPU 41 of the terminal control device 39. The control program stored in the first storage device 42 is executed repeatedly by the terminal CPU 41. Figure 3 The series of processes shown in section (B) is as follows.

[0062] exist Figure 3 In the series of processes shown in section (B), in step S31, the terminal CPU 41 determines whether synchronization with the vehicle control device 15 is established. If the terminal CPU 41 determines that synchronization with the vehicle control device 15 is established (S31: "Yes"), the process proceeds to step S33. On the other hand, if the terminal CPU 41 determines that synchronization with the vehicle control device 15 is not established (S31: "No"), the series of processes is temporarily terminated.

[0063] In step S33, the terminal CPU 41 obtains the vehicle model information sent from the vehicle control device 15. In step S35, the terminal CPU 41 starts recording via microphone 35. In step S37, the terminal CPU 41 begins obtaining the vehicle 10 status variables sent from the vehicle control device 15.

[0064] In step S39, the terminal CPU 41 determines whether there is a perception prompt, which indicates that the occupant of vehicle 10 has perceived an abnormal noise occurring in vehicle 10. For example, if the occupant performs a predetermined perception operation on portable terminal 30 as a pre-determined predetermined operation, the terminal CPU 41 considers the perception prompt to be present. Conversely, if the occupant does not perform the predetermined perception operation on portable terminal 30, the terminal CPU 41 considers the perception prompt to be absent. If the terminal CPU 41 determines that there is a perception prompt (S39: "Yes"), the process proceeds to step S41. On the other hand, if the terminal CPU 41 determines that there is no perception prompt (S39: "No"), the determination in step S39 is repeated until a perception prompt is determined to be present.

[0065] Here, Figure 4 The diagram illustrates an example of an abnormal noise occurring in vehicle 10. When such an abnormal noise occurs... Figure 4 In the case of the abnormal noise shown, the occupants of vehicle 10 sometimes feel uncomfortable with the noise. For example, in Figure 4 In this context, there are prominent peaks in the smooth curve representing the relationship between sound pressure level and frequency. In such cases, the occupant may sometimes perform a pre-defined sensing operation on the portable terminal 30.

[0066] Return to Figure 3 In part (B), in step S41, the terminal CPU 41 begins storing an audio signal, which is a signal related to the sound sensed by the microphone 35, and a state variable of the vehicle 10 obtained from the vehicle control device 15. At this time, the terminal CPU 41 stores the audio signal and the state variable together in the second storage device 43. That is, step S41 corresponds to the "audio signal acquisition processing". In step S43, the terminal CPU 41 determines whether a predetermined time has elapsed since the time point at which the aforementioned sensing prompt was determined. If the elapsed time has not elapsed (S43: "No"), the terminal CPU 41 returns the processing to step S41. That is, the terminal CPU 41 continues the processing of storing the audio signal and the state variable in the second storage device 43. On the other hand, if the elapsed time has elapsed (S43: "Yes"), the terminal CPU 41 transfers the processing to step S45.

[0067] In step S45, the terminal CPU 41 performs a transmission process. Specifically, during this process, the terminal CPU 41 transmits the time-series data of the sound signal stored in the second storage device 43 and the time-series data of the vehicle 10's state variables from the terminal communication unit 37 to the data analysis center 60. Furthermore, during the transmission process, the terminal CPU 41 transmits the vehicle model information obtained in step S33 and the model information of the microphone 35 of the portable terminal 30 from the terminal communication unit 37 to the data analysis center 60. After transmission is completed, the terminal CPU 41 temporarily terminates the series of processes.

[0068] Figure 5 as well as Figure 6 The diagram illustrates the processing flow executed by the central CPU 64 of the central control device 63. The central CPU 64 executes the control program stored in the first storage device 65 repeatedly. Figure 5 as well as Figure 6 The series of processes shown.

[0069] In a series of processes, in step S61, the central CPU 64 determines whether the central communication unit 61 has received the data sent by the portable terminal 30 to the data parsing center 60 in step S45 above. If the central communication unit 61 has received the data (S61: "Yes"), the central CPU 64 transfers the processing to step S63. On the other hand, if the central communication unit 61 has not received the data (S61: "No"), the central CPU 64 temporarily terminates the series of processes.

[0070] In step S63, the central CPU 64 obtains the model information of the microphone 35 received by the central communication unit 61. That is, step S63 corresponds to "model information acquisition and processing".

[0071] In step S65, the central CPU 64 obtains the vehicle model information of the vehicle 10 received by the central communication unit 61. In step S67, the central CPU 64 obtains the time-series data of the sound signal received by the central communication unit 61 and the time-series data of the state variables of the vehicle 10.

[0072] In step S69, the central CPU 64 determines whether the model of the microphone 35, as indicated by the model information obtained in step S63, is the same as the model of the learning microphone 35A. In this embodiment, the frequency characteristics of the learning microphone 35A are... Figure 2The frequency characteristic of microphone 35, as indicated by the model information, is "F characteristic". Therefore, if the frequency characteristic of microphone 35, as indicated by the model information, is "F characteristic", the central CPU 64 considers the model of microphone 35 to be the same as that of learning microphone 35A. On the other hand, if the frequency characteristic of microphone 35, as indicated by the model information, is not "F characteristic", the central CPU 64 considers the model of microphone 35 to be different from that of learning microphone 35A. Furthermore, if the central CPU 64 determines that the model of microphone 35 is the same as that of learning microphone 35A (S69: "Yes"), the process proceeds to step S71. On the other hand, if the central CPU 64 determines that the model of microphone 35 is different from that of learning microphone 35A (S69: "No"), the process proceeds to step S81.

[0073] In step S71, the central CPU 64 inputs the time-series data of the sound signal obtained in step S67 and the time-series data of the state variables of the vehicle 10 as input variables x into the mapping. Then, in step S73, the central CPU 64 obtains the output variable y from the mapping. That is, step S73 is a process of obtaining the output variable y from the mapping by inputting an uncorrected sound signal into the mapping when the model of the microphone 35 is the same as the model of the learning microphone 35A. Therefore, step S73 corresponds to the "reference variable acquisition process". The output variable y of step S73 corresponds to the reference variable.

[0074] After obtaining the output variable y through the central CPU 64 in step S73, the central CPU 64 transfers the processing to step S75. In step S75, the central CPU 64 determines the sound generation factor perceived by the microphone 35 based on the output variable y obtained in step S73. Specifically, the central CPU 64 selects the output variable with the largest value from M output variables y(1), y(2), ..., y(M). Furthermore, the central CPU 64 uses factor determination data 72 to determine the generation factor candidate corresponding to the selected output variable as the actual candidate. Therefore, step S75 corresponds to the "second factor determination processing". Moreover, the central CPU 64 transfers the processing to step S113.

[0075] In step S81, the central CPU 64 determines whether the frequency characteristics of the microphone 35 can be determined. For example, in Figure 2If the model data 73 includes a model that uses microphone model information, the central CPU 64 can determine the frequency characteristics of the microphone 35. Conversely, if the model data 73 does not include a model that uses microphone model information, the central CPU 64 cannot determine the frequency characteristics of the microphone 35. Furthermore, if the central CPU 64 determines that it can determine the frequency characteristics of the microphone 35 (S81: "Yes"), it transfers the process to step S83. Conversely, if the central CPU 64 determines that it cannot determine the frequency characteristics of the microphone 35 (S81: "No"), it transfers the process to step S91. That is, in... Figure 2 If the model information of microphone 35 is present in the model data 73, that is, if the model information of microphone 35 is stored in the second storage device 66, the central CPU 64 transfers the processing to step S83. On the other hand, if the model information of microphone 35 is not present in the model data 73, that is, if the model information of microphone 35 is not stored in the second storage device 66, the central CPU 64 transfers the processing to step S91.

[0076] In step S83, the central CPU 64 performs characteristic correction processing to make the frequency characteristics of the sound signal approximate the frequency characteristics of the learning sound signal by correcting the information corresponding to the model of the microphone 35. As will be described in detail later, the learning sound signal refers to the sound signal input to the mapping machine during the mapping process. The sound represented by the learning sound signal is the sound perceived by the learning microphone 35A. Therefore, in step S83, the central CPU 64 performs characteristic correction processing corresponding to the model information of the microphone 35. That is, if the model information of the microphone 35 is first model information, the central CPU 64 performs characteristic correction processing corresponding to the frequency characteristics of the microphone 35 represented by the first model information. On the other hand, if the model information of the microphone 35 is second model information, the central CPU 64 performs characteristic correction processing corresponding to the frequency characteristics of the microphone 35 represented by the second model information.

[0077] Here, an example of characteristic correction processing is explained. Let the frequency characteristics of the learning microphone 35A be such that its sensitivity to low-frequency sounds is relatively high, while its sensitivity to high-frequency sounds is relatively low. Conversely, let the frequency characteristics of the microphone 35 be such that its sensitivity to low-frequency sounds is relatively low, while its sensitivity to high-frequency sounds is relatively high. In this case, the frequency characteristics of the learning sound signal, like those of the learning microphone 35A, are characterized by high sensitivity to low-frequency sounds and low sensitivity to high-frequency sounds. Furthermore, the frequency characteristics of the sound signal related to the sound sensed by the microphone 35, like those of the microphone 35, are characterized by low sensitivity to low-frequency sounds and high sensitivity to high-frequency sounds. Therefore, in the characteristic correction processing, the central CPU 64 corrects the sound signal by increasing the sound pressure level in the low-frequency range and decreasing the sound pressure level in the high-frequency range. Thus, the central CPU 64 can make the frequency characteristics of the sound signal approximate the frequency characteristics of the learning sound signal.

[0078] In this embodiment, multiple characteristic correction processes are prepared in advance as characteristic correction processes. Therefore, when the model information of microphone 35 is the first model information, the central CPU 64 executes the first characteristic correction process as the characteristic correction process for the first model information. Furthermore, when the model information of microphone 35 is the second model information, the central CPU 64 executes the second characteristic correction process as the characteristic correction process for the second model information. The first characteristic correction process is the process that enables the frequency characteristics of the sound signal to approximate the frequency characteristics of the learning sound signal when the model information of microphone 35 is the first model information. The second characteristic correction process is the process that enables the frequency characteristics of the sound signal to approximate the frequency characteristics of the learning sound signal when the model information of microphone 35 is the second model information.

[0079] After correcting the sound signal through characteristic correction processing, the central CPU 64 transfers the processing to step S85. In step S85, the central CPU 64 inputs the time-series data of the corrected sound signal (which was corrected in step S83) and the time-series data of the vehicle 10's state variable obtained in step S67 as input variables xa into the mapping. In step S87, the central CPU 64 obtains the output variable y of the mapping. That is, step S87 corresponds to the "variable acquisition processing" which obtains the variable output from the mapping by inputting the sound signal corrected through characteristic correction processing into the mapping.

[0080] In step S89, the central CPU 64 performs a factor determination process to determine the source of the sound perceived by the microphone 35 based on the output variable y obtained in step S87. The processing content of step S89 is largely equivalent to that of step S75, so a detailed description is omitted. In this embodiment, step S89 corresponds to the "first factor determination process." After determining the source of the sound, the central CPU 64 transfers the processing to step S113.

[0081] Furthermore, in step S91, the central CPU 64 inputs the time-series data of the sound signal obtained in step S67 and the time-series data of the state variables of the vehicle 10 as input variables x into the mapping. That is, the central CPU 64 inputs the sound signal that has not been corrected through characteristic correction processing as input variable x into the mapping. In step S93, the central CPU 64 obtains the output variable y from the mapping. Step S93 corresponds to the "variable acquisition process" of obtaining the variable output from the mapping by inputting the uncorrected sound signal into the mapping. Moreover, the output variable y obtained in step S93 corresponds to the "third output variable".

[0082] After the output variable y is obtained in step S93, the central CPU 64 transfers the processing to step S95. In step S95, the central CPU 64 determines the source of the sound perceived by the microphone 35 based on the output variable y obtained in step S93. The processing content of step S95 is largely equivalent to that of step S75, so its detailed description is omitted.

[0083] In step S97, the central CPU 64 sets the counter F to 1. Furthermore, the central CPU 64 transfers the processing to step S99.

[0084] In step S99, the central CPU 64 performs characteristic correction processing corresponding to the counter F. For example, when the counter F is 1, the central CPU 64 performs characteristic correction processing Z(1) based on the premise that the frequency characteristic of the microphone 35 is "characteristic A". Additionally, for example, when the counter F is 2, the central CPU 64 performs characteristic correction processing Z(2) based on the premise that the frequency characteristic of the microphone 35 is "characteristic B". Furthermore, for example, when the counter F is 3, the central CPU 64 performs characteristic correction processing Z(3) based on the premise that the frequency characteristic of the microphone 35 is "characteristic A improvement". Furthermore, characteristic correction processing Z(1) is a characteristic correction process that, when the frequency characteristic of the microphone 35 is "characteristic A", makes the frequency characteristic of the sound signal closer to the frequency characteristic of the learning sound signal. Characteristic correction processing Z(2) is a characteristic correction process that, when the frequency characteristic of the microphone 35 is "characteristic B", makes the frequency characteristic of the sound signal closer to the frequency characteristic of the learning sound signal. Characteristic correction processing Z(3) is a characteristic correction process that makes the frequency characteristics of the sound signal close to the frequency characteristics of the learning sound signal when the frequency characteristics of the microphone 35 are "A characteristic improvement".

[0085] In step S101, the central CPU 64 inputs the time-series data of the corrected sound signal (which was corrected in step S99) and the time-series data of the vehicle 10's state variable (obtained in step S67) as input variables x(F) into the mapping. In step S103, the central CPU 64 obtains the mapping's output variable y. For example, when the characteristic correction process Z(1) is set to "first characteristic correction process", the mapping's output variable y when the counter F is 1 corresponds to "first output variable". Additionally, for example, when the characteristic correction process Z(2) is set to "second characteristic correction process", the mapping's output variable y when the counter F is 2 corresponds to "second output variable".

[0086] In step S105, the central CPU 64 determines the factors causing the sound sensed by the microphone 35 based on the output variable y obtained in step S103. The processing content of step S105 is largely the same as that of step S75, so its detailed description is omitted.

[0087] In step S107, the central CPU 64 increments the counter F by 1. In step S109, the central CPU 64 determines whether the counter F is greater than or equal to the determination value Fth. The determination value Fth is stored in... Figure 2 The data for model 73 shows the number of types of microphone frequency characteristics. Figure 2In the example shown, the microphone has 5 frequency characteristics, so the decision value Fth can be set to "5". If the counter F is greater than or equal to the decision value Fth (S109: "Yes"), the central CPU 64 transfers the process to step S111. Conversely, if the counter F is less than the decision value Fth (S109: "No"), the central CPU 64 transfers the process to step S99.

[0088] In step S111, the central CPU 64 performs a factor selection process to select the cause of the abnormal noise. That is, the central CPU 64 selects any one of the causes determined in step S95 and step S105. For example, the central CPU 64 selects the sound cause by taking a majority vote among the determined factors. After the selection of the cause of the abnormal noise is completed, the central CPU 64 transfers the processing to step S113.

[0089] In step S113, the central CPU 64 sends information related to the determined factors causing the sound from the central communication unit 61 to the portable terminal 30. Afterward, the central CPU 64 temporarily terminates a series of processes.

[0090] Furthermore, after obtaining information related to the cause of the sound sent by the data analysis center 60, the terminal CPU 41 of the terminal control device 39 notifies the passengers of the cause of the sound represented by the information. For example, the terminal CPU 41 displays the cause on the display screen 33.

[0091] <Learning Methods of Mapping>

[0092] Reference Figure 7 This describes a learning device 80 that performs machine learning on the mapping.

[0093] A learning sound signal, which is a signal related to the sound sensed by the learning microphone 35A, is input to the learning device 80. Additionally, a detection signal is input to the learning device 80 from the learning detection system 11A. One or more sensors constituting the learning detection system 11A are the same as one or more sensors constituting the detection system 11 of the vehicle 10.

[0094] The learning device 80 includes a learning CPU 81, a first storage device 82, and a second storage device 83. The first storage device 82 is a storage circuit that stores a control program executed by the learning CPU 81. The second storage device 83 is a storage circuit that stores mapping data 71a, which specifies the mapping that has not yet been completed in machine learning, and factor determination data 72.

[0095] Before performing machine learning on the mapping, the learning device 80 acquires multiple training data sets. These training data sets include the input variables for the mapping and the learning generation factors, which are the sound generation factors perceived by the learning microphone 35A. The input variables for the mapping include time-series data of the learning sound signals and time-series data of the vehicle 10's state variables.

[0096] The learning CPU 81 of the learning device 80 inputs the time-series data of the learning sound signal and the time-series data of the state variables contained in the training data into the mapping, and obtains the output variables y(1) to y(M) of the mapping. Next, the learning CPU 81 determines the sound generation factor based on the output variables y(1) to y(M) in the same way as in step S75 above. Moreover, the learning CPU 81 compares the determined sound generation factor with the learning generation factor contained in the training data. At this time, if the determined sound generation factor is different from the learning generation factor, the learning CPU 81 adjusts the various variables in the function approximator of the mapping in a way that increases the output variable corresponding to the learning generation factor among the output variables y(1) to y(M). For example, if the learning generation factor is a candidate for the first generation factor, the learning CPU 81 adjusts the various variables in the function approximator of the mapping in a way that maximizes the output variable y(1) among the output variables y(1) to y(M).

[0097] After the machine learning of such a mapping is completed, the mapping data 71 of the machine learning-processed mapping is stored in the second storage device 66 of the data parsing center 60.

[0098] <The function of this implementation method>

[0099] When microphone 35 detects an abnormal noise occurring in vehicle 10, the terminal CPU 41 of terminal control device 39 obtains an audio signal related to the sound detected by microphone 35. Furthermore, terminal control device 39 sends both the audio signal and the state variables of vehicle 10 to central control device 63. Additionally, the model information of microphone 35 is also sent to central control device 63.

[0100] The central CPU 64 of the central control device 63 corrects the frequency characteristics of the sound signal based on the acquired model information of the microphone 35. The model of the microphone 35 may differ from that of the learning microphone 35A (S69: "No"), but sometimes the model information of the microphone 35 is present in the model data 73 (S81: "Yes"). In this case, the central CPU 64 corrects the sound signal by performing characteristic correction processing corresponding to the model of the microphone 35, so that the frequency characteristics of the sound signal are close to the frequency characteristics of the learning sound signal. Next, the central CPU 64 determines the sound generation factor based on the output variable y output from the mapping by inputting the corrected sound signal to the mapping.

[0101] Furthermore, if the model of microphone 35 is the same as that of the learning microphone 35A (S69: "Yes"), the central CPU 64 inputs the uncorrected sound signal to the mapping. Moreover, the central CPU 64 determines the sound generation factor based on the output variable y output from the mapping.

[0102] On the other hand, if the model information for microphone 35 is not present in the model data 73 (S81: "No"), the central CPU 64 determines a candidate sound generation factor based on the output variable y output from the mapping by inputting an uncorrected sound signal to the mapping. This generation factor is referred to as "candidate factor Zr". Furthermore, the central CPU 64 repeatedly executes... Figure 6 The processing steps S99 to S109 shown determine Fth candidate factors for occurrence. Furthermore, the central CPU 64 determines the sound occurrence factor based on the candidate factors Zr and Fth candidates.

[0103] After determining the cause of the sound detected by the microphone 35 as described above, the central CPU 64 sends the determined cause information to the portable terminal 30. Then, the terminal CPU 41 of the terminal control device 39 notifies the owner of the portable terminal 30, i.e., the occupants of the vehicle 10, of the cause of the sound. The terminal CPU 41 uses predetermined hardware of the portable terminal 30, such as the display screen 33, vibration device, or sound device, to notify the occupants of the vehicle 10 of the cause of the sound.

[0104] <Effects of this implementation method>

[0105] (1-1) Even if the model of microphone 35 is different from that of learning microphone 35A (S69: "No"), if the model information of microphone 35 exists in model data 73 (S81: "Yes"), the sound signal is corrected by characteristic correction processing corresponding to that model (S83). As a result, the frequency characteristics of the sound signal input to the mapping are close to the frequency characteristics of the learning sound signal. As a result, the deviation of the frequency characteristics of the sound signal caused by the difference in the model of the microphone 35 that senses the sound can be reduced. Moreover, the sound generation factor sensed by microphone 35 is determined based on the output variable y output from the mapping by inputting the corrected sound signal into the mapping (S85-S89). As a result, the deviation in the accuracy of determining the sound generation factor corresponding to the model of microphone 35 can be reduced.

[0106] (1-2) The central control device 63 is capable of performing characteristic correction processing corresponding to various types of microphone models. Therefore, by determining the model of the microphone 35, the sound signal can be corrected through characteristic correction processing corresponding to that model. Moreover, the corrected sound signal is input to the mapping. Therefore, the effect of reducing the deviation in the accuracy of determining the sound generation factors corresponding to the model of the microphone 35 can be further improved.

[0107] (1-3) If the model of microphone 35 is the same as that of learning microphone 35A (S69: "Yes"), characteristic correction processing is not performed. Therefore, it is possible to suppress the unnecessary performance of characteristic correction processing, and correspondingly, it is possible to suppress the increase in processing load of the central CPU 64 of the central control device 63.

[0108] (1-4) If the model information for microphone 35 is not found in model data 73 (S81: "No"), by executing Figure 6 The processing steps S91 to S109 shown identify a large number of candidate sound generation factors. Furthermore, the sound generation factor is determined from these candidates. For example, the sound generation factor is determined by majority vote. Therefore, even if the model information for microphone 35 is not present in the model data 73, the accuracy of determining the sound generation factor can be suppressed.

[0109] (1-5) Model data 73 is stored in the second storage device 66 of the central control device 63. Additionally, it is executed by the central CPU 64 of the central control device 63. Figure 5 as well as Figure 6The process described is a series of steps. Therefore, when a new model of portable terminal is released, the model data 73 can be quickly updated. In addition, characteristic correction processing corresponding to the new microphone model can be easily prepared. Therefore, even if an abnormal noise is detected using the microphone of such a latest model of portable terminal, the accuracy of determining the cause of the sound can be improved.

[0110] Reference Figure 8 This document describes a second embodiment of a method and device for determining the causes of abnormal noise. Furthermore, in this second embodiment, the points where mapping data is stored in the storage device of the vehicle control device differ from those in the first embodiment. In the following description, the parts that differ from the first embodiment will be mainly explained; component structures identical to those in the first embodiment will be given the same reference numerals and repeated descriptions will be omitted.

[0111] Figure 8 The system shown includes a vehicle 10 and a portable terminal 30.

[0112] Vehicle 10 includes a detection system 11, a vehicle communication unit 13, and a vehicle control unit 15B. Vehicle control unit 15B includes a vehicle CPU 16, a first storage device 17, and a second storage device 18. The second storage device 18 pre-stores mapping data 71, factor determination data 72, and vehicle model data 73.

[0113] The portable terminal 30 includes a touch panel 31, a display screen 33, a microphone 35, a terminal communication device 37, and a terminal control device 39.

[0114] <Methods for Determining the Factors Causing Abnormal Noises>

[0115] exist Figure 8 In the system shown, mapping data 71, factor determination data 72, and model data 73 are stored in the second storage device 18 of the vehicle control device 15B. Therefore, the terminal CPU 41 of the terminal control device 39 sends the model information of the microphone 35 from the terminal communication device 37 to the vehicle control device 15B. Additionally, the terminal CPU 41 sends sound signals related to the sound sensed by the microphone 35 from the terminal communication device 37 to the vehicle control device 15B.

[0116] After receiving the audio signal from the terminal control device 39, the vehicle CPU 16 of the vehicle control device 15B executes the corresponding... Figure 5 as well as Figure 6 The processes shown in steps S69 to S113 are equivalent to the processes described above. That is, the vehicle CPU 16 of the vehicle control device 15B determines the factors that cause the sound.

[0117] Furthermore, in this embodiment, the vehicle control device 15B and the terminal control device 39 constitute an example of a "analysis device". The terminal CPU 41 of the terminal control device 39 and the vehicle CPU 16 of the vehicle control device 15B constitute an example of an "execution circuit of the analysis device". The terminal CPU 41 corresponds to the "first execution circuit", and the vehicle CPU 16 corresponds to the "second execution circuit". The second storage device 18 of the vehicle control device 15B corresponds to the "storage circuit of the analysis device". Furthermore, when the vehicle control device 15B is an example of an "abnormal noise generation factor determination device", the vehicle CPU 16 of the vehicle control device 15B corresponds to the "execution circuit of the abnormal noise generation factor determination device". The second storage device 18 of the vehicle control device 15B corresponds to the "storage circuit of the abnormal noise generation factor determination device".

[0118] <Effects of this implementation method>

[0119] In this embodiment, in addition to the effects equivalent to those of the first embodiment described above (1-1) to (1-4), the following effects can also be obtained.

[0120] (2-1) Even if the sound signal and the state variables of the vehicle 10 are not sent to the data analysis center 60 located outside the vehicle, the second embodiment can still determine the factors causing the sound perceived by the microphone 35. That is, even if the communication between the portable terminal 30 and the data analysis center 60 is unstable, the second embodiment can still determine the factors causing the sound.

[0121] (Example of Change)

[0122] The above-described embodiments can be modified as described below. The above-described embodiments and the following modifications can be combined and implemented within a technically compatible framework.

[0123] ■In the first embodiment, the characteristic correction processing is performed by the central CPU 64 of the central control device 63, but it is not limited to this. For example, the characteristic correction processing may also be performed by the terminal CPU 41 of the terminal control device 39, and the terminal CPU 41 sends the sound signal corrected by the characteristic correction processing to the central control device 63. In this case, it is preferable to store the model data 73 in the second storage device 43 of the terminal control device 39.

[0124] ■In the second embodiment, the characteristic correction processing is performed by the vehicle CPU 16 of the vehicle control device 15B, but it is not limited to this. For example, the characteristic correction processing may also be performed by the terminal CPU 41 of the terminal control device 39, and the terminal CPU 41 sends the sound signal corrected by the characteristic correction processing to the vehicle control device 15B. In this case, it is preferable to store the model data 73 in the second storage device 43 of the terminal control device 39.

[0125] ■ In the above-described embodiments, even if the model of microphone 35 is the same as that of learning microphone 35A, it is possible to achieve the same result by performing the same operation as... Figure 5 as well as Figure 6 The processing steps S91 to S111 shown are equivalent to the processing steps to determine the factors that cause the sound.

[0126] ■ In the above-described embodiments, even if the model information of microphone 35 exists in the model data 73, it can still be achieved by executing the above-described embodiments. Figure 5 as well as Figure 6 The processing steps S91 to S111 shown are equivalent to the processing steps to determine the factors that cause the sound.

[0127] ■In the above-described embodiments, even if the model information for microphone 35 is not present in the model data 73, the cause of sound generation can be determined based on the output variable y output from the mapping by inputting an uncorrected audio signal to the mapping. Furthermore, any one of the multiple characteristic correction processes can be set as a predetermined characteristic correction process. Moreover, even if the model information for microphone 35 is not present in the model data 73, the cause of sound generation can be determined based on the output variable y output from the mapping by inputting an audio signal corrected by the predetermined characteristic correction process to the mapping.

[0128] ■ In the first embodiment described above, the terminal control device 39 sends an audio signal and the state variables of the vehicle 10 to the central control device 63, but is not limited thereto. For example, the audio signal can also be sent from the terminal control device 39 to the vehicle control device 15, and the vehicle control device 15 can send the audio signal and the state variables to the central control device 63.

[0129] ■ In the above-described embodiments, modifications may also be made. Figure 6 The execution order of steps S91 to S109 is shown. For example, steps S97 to S109 can also be executed, and after the determination in step S109 becomes "yes", steps S91 to S95 can be executed.

[0130] ■ In the above-described embodiments, when the factors causing the sound sensed by the microphone 35 are determined, the determination result is communicated to the occupants of the vehicle 10 via the portable terminal 30, but this is not a limitation. For example, the determination result of the factors causing the sound may also be communicated to the occupants by using an in-vehicle device as predetermined hardware.

[0131] ■ In the above-described embodiments, when the factors causing the sound sensed by the microphone 35 are determined, it is not necessary to notify the occupants of the vehicle 10 of the determination result.

[0132] ■If a microphone is installed in the cabin of vehicle 10, the factors that cause the sound sensed by the microphone can also be determined.

[0133] Specifically, in the first embodiment described above, the vehicle CPU 16 of the vehicle control device 15 acquires an audio signal. Therefore, the vehicle CPU 16 sends the audio signal to the data parsing center 60. In this case, the vehicle control device 15 and the central control device 63 constitute an example of a "parsing device," and the vehicle CPU 16 of the vehicle control device 15 and the central CPU 64 of the central control device 63 constitute an example of an "execution circuit of the parsing device." Furthermore, the vehicle CPU 16 corresponds to the "first execution circuit," and the central CPU 64 corresponds to the "second execution circuit."

[0134] Even in the second embodiment described above, the vehicle CPU 16 of the vehicle control device 15B also acquires the sound signal. In this case, the vehicle control device 15B corresponds to the "analysis device", so the vehicle CPU 16 of the vehicle control device 15B corresponds to the "execution circuit of the analysis device".

[0135] ■ Neural networks are not limited to feedforward networks with only one intermediate layer. For example, neural networks can be networks with two or more intermediate layers, or they can be convolutional neural networks or recurrent neural networks.

[0136] ■ The learned model obtained using machine learning does not necessarily have to be a neural network. For example, a support vector machine can also be used as a learned model.

[0137] ■The central control unit 63, the terminal control unit 39, and the vehicle control units 15 and 15B are not limited to examples that have a CPU and ROM to perform software processing. That is, such control units can be any of the structures of (a) to (c) below.

[0138] (a) The control device includes one or more processors that perform various processes according to a computer program. The processor includes a CPU and memories such as RAM and ROM. The memories store program code or instructions that enable the CPU to perform processes. Memory, i.e., non-transitory computer-readable storage media, includes any usable medium that can be accessed by a general-purpose or special-purpose computer.

[0139] (b) The control device has one or more dedicated hardware circuits for performing various processes. Examples of dedicated hardware circuits include, for example, application-specific integrated circuits, i.e., ASICs or FPGAs. Furthermore, ASIC is short for "Application Specific Integrated Circuit," and FPGA is short for "Field Programmable Gate Array."

[0140] (c) The control device has a processor that performs a portion of the various processes according to a computer program, and dedicated hardware circuitry that performs the remaining processes in the various processes.

Claims

1. A method for determining the causes of abnormal noise, wherein, The method for determining the factors causing abnormal noise includes: The parsing device stores mapping data through its storage circuit. The mapping data defines a mapping that takes a sound signal, which is a signal related to the sound perceived by the microphone, as input and outputs variables related to the factors that cause sound in the vehicle. The mapping is obtained by performing machine learning. When performing machine learning on the mapping, the sound signal input to the mapping is a learning sound signal, and the microphone that perceives the sound represented by the learning sound signal is a learning microphone. The execution circuit of the analytical device performs sound signal acquisition processing, in which the sound signal related to the sound perceived by the microphone is acquired; The device model information is obtained through the execution circuit, and the device model information is related to the device model of the microphone; If the obtained model information is different from the model information of the learning microphone, the execution circuit performs a characteristic correction process. In this characteristic correction process, the sound signal obtained by the sound signal acquisition process is corrected according to the obtained model information, so that the frequency characteristics of the sound signal are close to the frequency characteristics of the learning sound signal. The execution circuit performs variable acquisition processing. In this variable acquisition processing, if the acquired model information is the same as the model information of the learning microphone, the variable is obtained by inputting the sound signal obtained in the sound signal acquisition processing into the mapping. If the acquired model information is different from the model information of the learning microphone, the variable is obtained by inputting the sound signal corrected in the characteristic correction processing into the mapping. as well as The execution circuit performs a factor determination process, in which the factors causing the sound perceived by the microphone are determined based on the variables obtained in the variable acquisition process.

2. The method for determining the causes of abnormal noise according to claim 1, wherein, The storage circuit stores information about multiple of the aforementioned models. The multiple model information includes first model information and second model information. The method for determining the causes of abnormal noise also includes: If the obtained model information is the first model information, the execution circuit performs the first characteristic correction process as the characteristic correction process. as well as If the obtained model information is the second model information, the execution circuit performs the second characteristic correction process as the characteristic correction process. In the first characteristic correction process, the obtained sound signal is corrected according to the first model information to make the frequency characteristics of the sound signal close to the frequency characteristics of the learning sound signal. In the second characteristic correction process, the sound signal obtained by correcting it according to the second model information is made so that the frequency characteristics of the sound signal are close to the frequency characteristics of the learning sound signal.

3. The method for determining the causes of abnormal noise according to claim 2, wherein, The method for determining the causes of abnormal noise also includes: If the obtained model information is not stored in the storage circuit, The execution circuit performs the first characteristic correction process and the second characteristic correction process. The execution circuit inputs the sound signal, which has been corrected by the first characteristic correction process, into the mapping, thereby obtaining the variable output from the mapping as the first output variable; The execution circuit inputs the sound signal, which has been corrected by the second characteristic correction process, into the mapping, thereby obtaining the variable output from the mapping as the second output variable; The execution circuit inputs the sound signal obtained through the sound signal acquisition and processing to the mapping, thereby obtaining the variable output from the mapping as the third output variable; and The execution circuit performs a factor selection process, in which the sound generation factor is selected from the sound generation factors based on the first output variable, the sound generation factors based on the second output variable, and the sound generation factors based on the third output variable.

4. The method for determining the causes of abnormal noise according to any one of claims 1 to 3, wherein, The factor determination process is the first factor determination process. If the model of the microphone shown in the obtained model information is the same as the model of the learning microphone, The method for determining the causes of abnormal noise also includes: The execution circuit performs a reference variable acquisition process, in which the sound signal obtained by the sound signal acquisition process is input to the mapping, thereby obtaining the variable output from the mapping as the reference variable; as well as The execution circuit performs a second factor determination process, in which the generation factor of the sound perceived by the microphone is determined based on the reference variable obtained through the reference variable acquisition process.

5. The method for determining the causes of abnormal noise according to any one of claims 1 to 3, wherein, The execution circuit includes: a first execution circuit disposed in the vehicle or in a portable terminal held by an occupant of the vehicle; and a second execution circuit disposed outside the vehicle. The characteristic correction process, the variable acquisition process, and the factor determination process are performed by the second execution circuit.

6. A method for determining the causes of abnormal noise, the method comprising: The parsing device stores mapping data through its storage circuit. The mapping data defines a mapping that takes a sound signal, which is a signal related to the sound perceived by the microphone, as input and outputs variables related to the factors that cause sound in the vehicle. The mapping is obtained by performing machine learning. When performing machine learning on the mapping, the sound signal input to the mapping is a learning sound signal, and the microphone that perceives the sound represented by the learning sound signal is a learning microphone. The execution circuit of the analytical device performs sound signal acquisition processing, in which the sound signal related to the sound perceived by the microphone is acquired; The device model information is obtained through the execution circuit, and the device model information is information related to the device model of the microphone; The execution circuit performs a first characteristic correction process to correct the frequency characteristics of the sound signal obtained through the sound signal acquisition process. In the case that the model information of the microphone is a first model information that is different from the model information of the learning microphone, the first characteristic correction process can make the frequency characteristics of the sound signal close to the frequency characteristics of the learning sound signal. The execution circuit performs a second characteristic correction process to correct the frequency characteristics of the sound signal obtained through the sound signal acquisition process. In the case that the model information of the microphone is a second model information that is different from the model information of the learning microphone, the second characteristic correction process can make the frequency characteristics of the sound signal close to the frequency characteristics of the learning sound signal. The execution circuit performs a variable acquisition process, in which the variable acquisition process obtains a first output variable by inputting the sound signal corrected by the first characteristic correction process into the mapping, and obtains a variable output from the mapping as a first output variable by inputting the sound signal corrected by the second characteristic correction process into the mapping, and obtains a second output variable by inputting the sound signal obtained by the sound signal acquisition process into the mapping, and obtains a third output variable by inputting the variable output from the mapping; and The execution circuit performs a factor selection process, in which the sound generation factor is selected from the sound generation factors based on the first output variable, the sound generation factors based on the second output variable, and the sound generation factors based on the third output variable.

7. The method for determining the causes of abnormal noise according to claim 6, wherein, The execution circuit includes: a first execution circuit disposed in the vehicle or in a portable terminal held by an occupant of the vehicle; and a second execution circuit disposed outside the vehicle. The second execution circuit performs the first characteristic correction process, the second characteristic correction process, the variable acquisition process, and the factor selection process.

8. A device for determining the causes of abnormal noise, comprising determining the causes of sound sensed by a microphone, wherein, The device for determining the cause of abnormal noise includes an execution circuit and a storage circuit. The storage circuit stores mapping data with a predetermined mapping. This mapping takes as input an audio signal that is related to the sound perceived by the microphone, and outputs variables related to the factors that cause sound in the vehicle. The mapping is obtained by performing machine learning. When machine learning is applied to the mapping, the sound signal input to the mapping is a learning sound signal, and the microphone that perceives the sound represented by the learning sound signal is a learning microphone. The execution circuit is configured to execute: In the case where the model information of the microphone is different from the model information of the learning microphone, the frequency characteristics of the sound signal related to the sound perceived by the microphone are made close to the frequency characteristics of the learning sound signal by means of correction corresponding to the model information, which is information related to the model of the microphone. In the variable acquisition process, if the model information of the microphone is the same as that of the learning microphone, the variable output from the mapping is obtained by inputting the sound signal related to the sound perceived by the microphone into the mapping. If the model information of the microphone is different from that of the learning microphone, the sound signal corrected by the characteristic correction process is input into the mapping to obtain the variable output from the mapping. as well as The factor determination process determines the factors that cause the sound perceived by the microphone based on the variables obtained through the variable acquisition process.

9. A device for determining the causes of abnormal noise, comprising determining the causes of sound perceived by a microphone, wherein, The device for determining the cause of abnormal noise includes an execution circuit and a storage circuit. The storage circuit stores mapping data with a predetermined mapping, which takes a sound signal as input (as a signal related to the sound perceived by the microphone) and outputs variables related to the factors causing sound in the vehicle. The mapping is obtained by performing machine learning. When machine learning is applied to the mapping, the sound signal input to the mapping is a learning sound signal, and the microphone that perceives the sound represented by the learning sound signal is a learning microphone. The execution circuit is configured to execute: The first characteristic correction process corrects the frequency characteristics of the sound signal related to the sound perceived by the microphone, and when the model information of the microphone is a first model information that is different from the model information of the learning microphone, the first characteristic correction process can make the frequency characteristics of the sound signal close to the frequency characteristics of the learning sound signal. The second characteristic correction process corrects the frequency characteristics of the sound signal, and when the model information of the microphone is a second model information that is different from the model information of the learning microphone, the second characteristic correction process can make the frequency characteristics of the sound signal close to the frequency characteristics of the learning sound signal. The variable acquisition process involves inputting the sound signal corrected by the first characteristic correction process into the mapping to obtain a variable output from the mapping as a first output variable; inputting the sound signal corrected by the second characteristic correction process into the mapping to obtain a variable output from the mapping as a second output variable; and inputting the uncorrected sound signal into the mapping to obtain a variable output from the mapping as a third output variable. The factor selection process selects the sound occurrence factor from the sound occurrence factors based on the first output variable, the sound occurrence factors based on the second output variable, and the sound occurrence factors based on the third output variable.

Images