Accident detection device, accident detection method, and accident detection program

Abstract

To properly detect accidents.SOLUTION: An accident detection device 100B has an audio detection part 121, an impact detection part 124, a calculation part 126, and an accident detection part 125. The audio detection part 121 detects sounds and voices in a vehicle. The impact detection part 124 detects an impact occurring in the vehicle. The calculation part 126 calculates an average value of a first audio value detected during a first period. The accident detection part 125 detects an accident of the vehicle by using sounds and voices obtained by subtracting the average value during the first period from sounds and voices acquired during a second period including the timing when an impact of a predetermined value or more is detected.SELECTED DRAWING: Figure 14

Description

[Technical Field]

[0001] The present invention relates to an accident detection device, an accident detection method, and an accident detection program. [Background technology]

[0002] Drive recorders, which store vehicle status information (forward footage, vehicle speed, sudden acceleration / deceleration, etc.) before and after a car accident, provide useful information for investigating car collisions and other accidents. Therefore, they are increasingly being installed in transportation vehicles such as trucks, as well as commercial vehicles such as taxis and buses, and are also increasingly being installed in general vehicles. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2022-000818 Summary of the Invention [Problem to be solved by the invention]

[0004] However, conventional technologies may not properly detect accidents. For example, accidents may be determined based on the level of impact a vehicle receives and an emergency call may be made. However, conventional technologies may detect an accident as a sudden movement rather than an accident because the level of impact is low. Therefore, with conventional technologies, even though the vehicle occupants may recognize that an accident has occurred, the device may not determine that an accident has occurred, resulting in a situation where an emergency call is not made or the call cannot be connected.

[0005] The present invention has been made in view of the above, and has an object to provide an accident detection device, an accident detection method, and an accident detection program that are capable of appropriately detecting an accident. [Means for solving the problem]

[0006] In order to solve the above-mentioned problems and achieve the object, the accident detection device of the present invention is characterized by having a sound detection unit that detects sound inside the vehicle, an impact detection unit that detects an impact occurring to the vehicle, a calculation unit that calculates the average value of a first sound value detected in a first period, and an accident detection unit that detects a vehicle accident using sound obtained by subtracting the average value of the first period from sound obtained within a second period that includes the timing when an impact equal to or greater than a predetermined value is detected.

[0007] In addition, the accident detection method of the present invention is an accident detection method executed by an accident detection device, and is characterized by including: a sound detection process for detecting sound inside the vehicle; an impact detection process for detecting an impact occurring to the vehicle; a calculation process for calculating an average value of a first sound value detected in a first period; and an accident detection process for detecting a vehicle accident using sound obtained by subtracting the average value of the first period from sound obtained within a second period including the timing when an impact equal to or greater than a predetermined value is detected.

[0008] In addition, the accident detection program of the present invention is characterized in that it causes a computer to execute a sound detection procedure for detecting sound inside a vehicle, an impact detection procedure for detecting an impact occurring to the vehicle, a calculation procedure for calculating an average value of a first sound value detected in a first period, and an accident detection procedure for detecting a vehicle accident using sound obtained by subtracting the average value of the first period from sound obtained within a second period including the time when an impact equal to or greater than a predetermined value is detected. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 is a diagram showing an example of the overall configuration of an accident detection system common to all embodiments. [Figure 2] FIG. 2 is a diagram illustrating a process for detecting an accident using sound in addition to an impact. [Figure 3] FIG. 3 is a diagram illustrating the processing contents of the accident detection device according to the first embodiment. [Figure 4] FIG. 4 is a block diagram illustrating an example of the configuration of the accident detection device according to the first embodiment. [Figure 5] FIG. 5 is a diagram showing an example of audio data subjected to FFT analysis according to the first embodiment. [Figure 6] FIG. 6 is a diagram showing an example of audio data after FFT analysis used for accident detection according to the first embodiment. [Figure 7] FIG. 7 is a diagram illustrating an example of processing contents of the accident detection unit according to the first embodiment. [Figure 8] FIG. 8 is a diagram illustrating an example of processing contents of the accident detection unit according to the first embodiment. [Figure 9] FIG. 9 is a diagram illustrating an example of processing contents of the accident detection unit according to the first embodiment. [Figure 10] FIG. 10 is a diagram illustrating an example of processing performed by the change unit according to the first embodiment. [Figure 11] FIG. 11 is a diagram illustrating an example of processing performed by the accident detection unit when the score according to the first embodiment is used. [Figure 12] FIG. 12 is a flowchart illustrating an example of the flow of the accident detection process according to the first embodiment. [Figure 13] FIG. 13 is a diagram illustrating the processing contents of the accident detection device according to the second embodiment. [Figure 14] FIG. 14 is a block diagram illustrating an example of the configuration of an accident detection device according to the second embodiment. [Figure 15] FIG. 15 is a diagram illustrating a specific example of processing performed by the accident detection unit according to the second embodiment. [Figure 16] FIG. 16 is a flowchart showing an example of the flow of the accident detection process according to the second embodiment. [Figure 17] FIG. 17 is a diagram illustrating the processing contents of the accident detection device according to the third embodiment. [Figure 18] FIG. 18 is a flowchart showing an example of the flow of the accident detection process according to the third embodiment. [Figure 19] FIG. 19 is a hardware configuration diagram showing an example of a computer that realizes the functions of the accident detection devices according to the first, second, and third embodiments. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, first, second, and third embodiments of an accident detection device, an accident detection method, and an accident detection program according to the present invention will be described in detail with reference to the drawings. Note that the present invention is not limited to the first, second, and third embodiments described below.

[0011] [Introduction] Fig. 1 is a diagram showing an example of the overall configuration of an accident detection system common to all embodiments. As shown in Fig. 1, the accident detection system according to the present invention includes an accident detection device 100 mounted on a vehicle and an external server device 200. The accident detection device 100 appropriately detects vehicle accidents using image data captured while the vehicle is traveling.

[0012] The accident detection device 100 is realized by a computer or the like installed in a vehicle. For example, the accident detection device 100 may be a dedicated navigation device built into or mounted on a vehicle. For example, the accident detection device 100 may be configured with a navigation device and a recording device (drive recorder). As one example, the accident detection device 100 may be a composite device in which a navigation device and a recording device that are independent from each other are connected so that they can communicate with each other. As another example, the accident detection device 100 may be a single device having a navigation function and a recording function.

[0013] The accident detection device 100 may also be configured with a sensor device and a communication device. As one example, the accident detection device 100 may be a composite device in which a sensor device and a notification device that are independent of each other are connected to each other so that they can communicate with each other. As another example, the accident detection device 100 may be a single device that has a sensor function and a notification function.

[0014] Furthermore, a vehicle occupant can connect a predetermined sensor to a portable terminal device (for example, a smartphone, tablet terminal, notebook PC, desktop PC, PDA, etc.) that they use on a daily basis and install a predetermined application, thereby substituting the device as the accident detection device 100. For example, a portable terminal device that is equipped with a predetermined sensor or to which a predetermined sensor is connected can be understood as the accident detection device referred to here. When a portable terminal device is used as the accident detection device 100, it is installed, for example, on the dashboard of the vehicle while driving.

[0015] When the detection processing unit 120 detects that there is a possibility of an accident, the external server device 200 is notified of an event that an accident has occurred. For example, when the detection processing unit 120 notifies the external server device 200 of an event that an accident has occurred, the external server device 200 establishes communication between the vehicle occupant and a call center so that the vehicle occupant can make a call to the call center, which is the destination of an emergency call.

[0016] However, a typical accident determination device detects accidents using only impacts, and may not properly detect accidents. For example, when a vehicle receives an impact, if the level of the impact is analyzed to determine whether or not an accident has occurred, the impact level may be so low that it is not detected as an accident, but may be detected as a sudden movement. As a result, although the driver of the vehicle may recognize an accident, the accident determination device may not determine an accident, and therefore an accident report may not be sent to the external server device 200, resulting in no call to the call center. In addition, an erroneous determination may be made that an accident has occurred when no accident has actually occurred.

[0017] Therefore, as shown in FIG. 1, the accident detection device 100 according to the embodiment analyzes not only the impact that occurred on the vehicle but also the audio data before and after the impact was detected, so that an accident that was erroneously determined to be caused by a sudden movement (sudden acceleration, sudden braking, sudden steering, etc.) because the detected impact level was low can be properly determined to be an accident based on the volume peak at the time of the sudden movement.

[0018] However, even in accident detection methods that use both impact and sound, there is a possibility that the accuracy may decrease due to false detection of an accident. Therefore, the following describes in detail the circumstances in which false detection of an accident may occur.

[0019] Using Fig. 2, we will explain a technology executed by the accident detection device 100 that detects accidents by combining an impact that has occurred on a vehicle with the sound that was generated at the time, and an example of a false detection. Fig. 2 is a diagram that explains the processing when accident detection is performed using sound in addition to an impact. Fig. 2 shows an example of determining whether a vehicle accident has occurred by determining whether the volume peak when an impact that has occurred on a vehicle is detected exceeds a preset volume threshold.

[0020] In this case, the vehicle accident detection device detects the sound inside the vehicle and uses it for accident determination processing, so it is affected by noise from outside the vehicle, such as road construction, and noise from inside the vehicle, such as a loud car stereo, etc. In other words, the sound inside the vehicle other than when the impact is detected is amplified by the noise, and the volume peak of the sound data other than when the impact is detected exceeds the volume threshold.

[0021] As a result, it is not possible to properly determine the contact sound that occurs when an accident occurs from the audio inside the vehicle, so there is a possibility of false detection when detecting a vehicle accident using audio data at the time of impact detection.

[0022] Therefore, the accident detection device 100 of this embodiment detects the normal noise level inside the vehicle from audio data other than when an impact is detected, and changes the volume threshold or corrects the audio data according to the noise level, thereby making it possible to properly detect a vehicle accident from audio data when an impact is detected.

[0023] In the first embodiment, a process for changing a volume threshold value according to the loudness of noise will be described, and in the second embodiment, a process for correcting a value of audio data according to the loudness of noise will be described. The accident detection device 100A according to the first embodiment and the accident detection device 100B according to the second embodiment are each an example of the accident detection device 100.

[0024] [Embodiment 1] (Processing contents of the accident detection device 100A) The processing details of the accident detection device 100A according to the first embodiment will be described. The accident detection device 100A according to the first embodiment determines the noise state inside the vehicle and detects a vehicle accident based on an impact occurring to the vehicle and sound inside the vehicle. When detecting an accident, the accident detection device 100A changes the threshold value of the sound used to detect an accident, or changes the weight of the sound when detecting an accident by combining impact and sound, based on the determination result of the noise state inside the vehicle.

[0025] Here, the processing contents of the accident detection device 100A will be described with a specific example. Fig. 3 is a diagram illustrating the processing contents of the accident detection device according to the first embodiment. Fig. 3 shows a processing example in which a preset volume threshold is changed based on the state of noise inside the vehicle.

[0026] For example, if the overall volume of audio data detected during normal driving is high due to noise from outside the vehicle or noise from inside the vehicle, the accident detection device 100A determines that the vehicle interior is in a noisy state. Since the vehicle interior is in a noisy state, the accident detection device 100A raises the volume threshold used to determine whether the detected audio is a contact sound related to an accident, adjusting it so that only the volume peak of the contact sound detected during an accident exceeds the volume threshold. When the accident detection device 100A detects an impact on the vehicle as a sudden movement, the volume peak near the point where the sudden movement was detected exceeds the volume threshold, thereby detecting a vehicle accident.

[0027] As a result, even if the impact on the vehicle is detected as a sudden movement that is less than an accident, the accident detection device 100A can properly detect a vehicle accident because only the audio data at the time of the sudden movement detection exceeds the adjusted volume threshold.

[0028] (Functional configuration of the accident detection device 100A) Next, an example of the functional configuration of the accident detection device 100A according to embodiment 1 will be described. Fig. 4 is a block diagram showing an example of the configuration of the accident detection device according to embodiment 1. As shown in Fig. 4, the accident detection device 100A includes, for example, a storage unit 110, a detection processing unit 120, a camera 130, a microphone 140, a video encoder processing unit 150, an audio encoder processing unit 160, a muxer unit 170, and an acceleration sensor 180.

[0029] Here, the detection processing unit 120, the video encoder processing unit 150, the audio encoder processing unit 160, and the muxer unit 170 are realized by a control unit such as a processor. The control unit such as a processor has an internal memory for storing programs that define various processing procedures and required data, and executes various processes using these. Here, the control unit is, for example, an electronic circuit such as a CPU (Central Processing Unit) or an MPU (Micro Processing Unit), or an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array). Furthermore, each unit may be realized by a processor.

[0030] The video encoder processing unit 150 is a processing unit that converts analog image data captured by the camera 130 into a digital signal. The audio encoder processing unit 160 is software that converts the format of an audio file of audio data acquired by the microphone 140.

[0031] Acceleration sensor 180 measures the acceleration of the vehicle. For example, acceleration sensor 180 measures acceleration along three axes: X-axis, Y-axis, and Z-axis. Acceleration sensor 180 then outputs information about the measured acceleration of the vehicle to detection processing unit 120.

[0032] The muxer unit 170 combines the imaging data and audio data and outputs the moving image file 111. For example, the muxer unit 170 combines the imaging data output from the video encoder processing unit 150 and the audio data output from the audio encoder processing unit 160 and outputs the moving image file 111.

[0033] The memory unit 110 stores various programs executed by the detection processing unit 120, data required when the detection processing unit 120 performs processing, video files 111 output by the muxer unit 170, analyzed audio data output by the audio detection unit 121 described later, and the like.

[0034] The detection processing unit 120 detects a vehicle accident from the acquired voice data, measurement data of the acceleration sensor 180, etc. The detection processing unit 120 has a voice detection unit 121, a determination unit 122, a change unit 123, an impact detection unit 124, and an accident detection unit 125.

[0035] The voice detection unit 121 performs a short-time Fourier transform (STFT) on the voice data acquired by the microphone 140 to generate analysis data indicating the sum of the amplitudes of all frequency bands of the voice data for each time period, and stores the analysis data in the storage unit 110. For example, the voice detection unit 121 sequentially performs processing on the voice data by a windowing unit 121a and an FFT unit 121b, which will be described later.

[0036] The windowing unit 121a is a processing unit that multiplies input signal data, which is audio data acquired from the microphone 140 or the like, by a window function of a specific analysis frame length on the time axis. For example, the windowing unit 121a extracts frames of a specific time length from the input signal data for each frame period and multiplies the frames by a window function, for example, a Hanning window.

[0037] To reduce information loss due to the window function, the windowing unit 121a can overlap the preceding and following analysis frames at any desired rate. For example, by analyzing a fixed frame length of 512 samples at a fixed frame period of 256 samples, the overlap rate can be set to 50%. The analysis frame obtained in this way is output to the FFT unit 121b.

[0038] The FFT unit 121b is a processing unit that executes a Fast Fourier Transform (FFT). For example, the FFT unit 121b applies an FFT to the analysis frame that has been windowed by the windowing unit 121a. This converts the input signal of the analysis frame into an amplitude spectrum and a phase spectrum.

[0039] Thereafter, FFT unit 121b stores the amplitude spectrum obtained by FFT in storage unit 110. Furthermore, FFT unit 121b calculates the sum of the amplitudes of all frequency bands of the amplitude spectrum obtained by FFT, and stores the sum in storage unit 110. Note that although an example in which FFT is applied has been given here, other algorithms such as Fourier transform and discrete Fourier transform may also be applied to convert from the time domain to the frequency domain.

[0040] Here, the difference between the sound data after FFT analysis in normal times and when a contact sound is generated will be described. Fig. 5 is a diagram showing an example of sound data that has undergone FFT analysis according to embodiment 1. In the example of Fig. 5, an FFT is performed on sound data that includes a contact sound acquired by the microphone 140, and the amplitude value is shown for each frequency band.

[0041] As shown in Figure 5, the graph showing the FFT results for normal times when no contact sound is included shows large amplitude values ​​in the low frequency band, indicating that the overall amplitude values ​​are small. On the other hand, the graph showing the FFT results for when contact sound is present also shows large amplitude values ​​in the high frequency band, indicating that the overall amplitude values ​​are large.

[0042] Next, a description will be given of analysis data obtained by calculating the sum of amplitudes of all frequency bands after performing FFT on the audio data including the contact sound shown in Fig. 5. Fig. 6 is a diagram showing an example of audio data after FFT analysis used for accident detection according to embodiment 1. The example of Fig. 6 shows a graph in which the sum of amplitudes of all frequency bands after performing FFT on the audio data including the contact sound shown in Fig. 5 is calculated for each playback time (seconds).

[0043] In the graph shown in Figure 6, the vertical axis represents the sum of the amplitudes of all frequency bands (Σ), and the horizontal axis represents the playback time (seconds) of the acquired audio data. As shown in Figure 6, the total amplitude value exceeds 600 around 9.7 seconds when the contact sound occurred, while the maximum total amplitude value at other times remains around 300. In other words, over the 16-second playback time, the audio data around 9.7 seconds when the contact sound occurred due to the accident is detected as a volume peak.

[0044] As a result, the audio detection unit 121 can generate analysis data from the audio data acquired by the microphone 140, which is used by the accident detection unit 125 (described later) and is capable of detecting volume peaks.

[0045] The determination unit 122 determines the noise state inside the vehicle. For example, the determination unit 122 calculates the volume of the sound during normal driving from the sound data detected by the sound detection unit 121, and determines whether the current state inside the vehicle is a noisy state, a normal state, or a quiet state.

[0046] For example, the determination unit 122 calculates the average value of the total amplitude value over a predetermined time period from the analyzed audio data generated by the audio detection unit 121, and sets this as the audio value during normal driving. Then, the determination unit 122 determines the state of noise inside the vehicle as a noisy state when the audio value during normal driving is 300 or more, a normal state when the audio value during normal driving is less than 300 and more than 100, and a quiet state when the audio value during normal driving is 100 or less.

[0047] The change unit 123 changes the threshold value of the sound used to detect an accident, or changes the weight of the sound when detecting an accident by combining an impact and sound, based on the determination result of the noise state inside the vehicle. A specific example of the threshold or weight change process performed by the change unit 123 will be described below.

[0048] For example, when changing the audio threshold and when the determination unit 122 determines that the noise inside the vehicle is quiet, the change unit 123 changes the threshold set for determining whether the volume peak is a contact sound to a value smaller than the set value.

[0049] In addition, when changing the audio threshold and when the determination unit 122 determines that the noise inside the vehicle is in a noisy state, the change unit 123 changes the threshold set for determining whether the volume peak is a contact sound to a value greater than the set value.

[0050] Furthermore, when changing the weight of the voice, the change unit 123 reduces the weight of the voice as the noise inside the vehicle determined by the determination unit 122 increases. For example, when the noise inside the vehicle is in a noisy state, the change unit 123 determines that the reliability of the voice detection result is low, and increases the weight of the impact detection result and reduces the weight of the voice detection result.

[0051] The impact detection unit 124 detects an impact that has occurred to the vehicle. For example, the impact detection unit 124 calculates a resultant vector and a moving average using measurement data of acceleration values ​​of three axes, i.e., the X-axis, Y-axis, and Z-axis, acquired by the acceleration sensor 180, and notifies the accident detection unit 125 of the calculated values ​​as the impact detection result.

[0052] The accident detection unit 125 detects a vehicle accident based on the impact on the vehicle and the sound inside the vehicle. For example, the accident detection unit 125 uses the resultant vector and moving average data of the impact detected by the impact detection unit 124 and the analyzed sound data generated by the sound detection unit 121, and compares the numerical value of each data with a preset threshold value for each data, thereby detecting a vehicle accident.

[0053] First, we will explain the process of determining whether an accident is a major accident, a minor or medium accident, or a sudden behavior based on the detection result of an impact to the vehicle by the accident detection unit 125. For example, the accident detection unit 125 detects a major accident when the resultant vector calculated by the impact detection unit 124 is equal to or greater than the major accident threshold, and detects a minor or medium accident when the resultant vector is smaller than the major accident threshold and equal to or greater than the minor or medium accident threshold.

[0054] On the other hand, for example, when the resultant vector is less than the small to medium accident threshold, the accident detection unit 125 detects sudden behavior such as sudden deceleration, sudden acceleration, sudden right steering, or sudden left steering by comparing the moving average with a threshold.

[0055] Here, when the accident detection unit 125 detects a sudden movement in which the detected impact is less than a small to medium accident, for example, the accident detection unit 125 determines whether the sudden movement is an accident by determining whether conditions 1 to 3 described below are satisfied. The determination contents of conditions 1 to 3 will be explained below in order.

[0056] (Explanation of condition 1) Condition 1 is a condition for determining whether a volume deviating from normal volume has been detected from audio data after analysis for about 20 seconds including the time point at which an impact was detected. Fig. 7 is a diagram showing an example of the processing content of the accident detection unit according to the first embodiment. As shown in Fig. 7, the accident detection unit 125 detects the volume peak at around 9.7 seconds, which indicates a contact sound, as audio that may be deviating from normal volume and may indicate a contact.

[0057] Here, a specific example of the determination content of condition 1 by the accident detection unit 125 will be described. Fig. 8 is a diagram showing an example of the processing content of the accident detection unit according to embodiment 1. Fig. 8 shows an example of the case where the sum of the amplitudes of all frequency bands of audio data is calculated at intervals of 100 ms by FFT, and a sudden change in volume is detected when the calculated sum of amplitudes satisfies the following (1) to (4):

[0058] First, the accident detection unit 125 performs outlier determination to determine whether the target amplitude is an outlier compared to the immediately preceding sound field (Condition 1-(1)). Specifically, the accident detection unit 125 determines whether the following condition is satisfied: "total amplitude value Σ of the target amplitude for 100 ms > average amplitude Σ for the immediately preceding 5 seconds + 3σ." Note that σ is the standard deviation.

[0059] Next, the accident detection unit 125 performs a gradient determination to determine whether the target amplitude has a sudden volume change (Condition 1-(2)). Specifically, the accident detection unit 125 determines whether the following is satisfied: "sum of amplitudes Σ of the target amplitude of 100 ms - sum of amplitudes Σ of the immediately preceding amplitude of 100 ms > gradient threshold."

[0060] Next, the accident detection unit 125 performs a contact volume determination to determine whether the target amplitude is equivalent to the contact volume (Condition 1-(3)). Specifically, the accident detection unit 125 determines whether the "total amplitude value Σ of the target amplitude of 100 ms > contact volume threshold" is satisfied.

[0061] Then, the accident detection unit 125 performs a detection validity determination to determine whether or not the immediately preceding sound field is in a detectable state (Condition 1-(4)). Specifically, the accident detection unit 125 determines whether or not "amplitude σ for the immediately preceding 5 seconds < detection validity σ threshold" is satisfied.

[0062] 8, the accident detection unit 125 performs a determination of the above-described conditions 1-(1) to 1-(4) and thereby detects the sum of the amplitudes around 9.7 seconds as a volume peak that may be a contact sound that deviates from normal conditions and satisfies all of conditions 1-(1) to 1-(4). In other words, the accident detection unit 125 determines that condition 1 is satisfied. Note that the order in which the accident detection unit 125 performs a determination of the above-described conditions 1-(1) to 1-(4) is not particularly limited.

[0063] (Explanation of condition 2) Condition 2 is a condition for determining whether or not a volume peak that satisfies the above-mentioned condition 1 exists between 2 seconds before and 1 second after the point in time when a sudden movement is detected. Fig. 9 is a diagram showing an example of the processing content of the accident detection unit according to the first embodiment. Fig. 9 shows an example in which a sudden movement is detected at a point in time of 10 seconds of playback time in the audio data.

[0064] As shown in FIG. 9, the accident detection unit 125 determines that condition 2 is met because a volume peak that satisfies the above-mentioned condition 1 is detected around 9.7 seconds, between 10 seconds into the playback time when the sudden behavior was detected and the previous 2 seconds.

[0065] (Explanation of condition 3) Condition 3 is a condition for determining whether the vehicle slows down or stops within a certain distance from the point where the sudden movement is detected. The certain distance is the distance at which the vehicle is expected to slow down or stop in the event of a vehicle accident, and is set to, for example, 100 m. Note that the certain distance is not limited to 100 m, but may be 200 m, and can be changed as appropriate based on the results of demonstrations and research.

[0066] The certain distance can be acquired by a GPS (Global Positioning System) sensor (not shown). Note that the method for measuring the distance is not limited to a GPS sensor, and various methods capable of measuring the distance can be used. For example, the distance may be measured based on parameters acquired from a vehicle speed sensor, or based on an odometer value acquired via a CAN (Controller Area Network).

[0067] For example, if the vehicle speed becomes equal to or lower than a slow speed within 100 m, which is an example of a certain distance, from the point where the sudden behavior was detected, the accident detection unit 125 determines that condition 3 is satisfied. Explaining this using the example of Fig. 9, the accident detection unit 125 uses the vehicle position at the point 10 seconds into the playback time when the sudden behavior was detected as a reference, and determines that condition 3 is satisfied if the vehicle speed within a certain distance (e.g., 100 m) from the reference becomes equal to or lower than a threshold value (e.g., 5 km / hour) or if the vehicle speed becomes 0 m / hour.

[0068] Furthermore, condition 3 is not limited to the distance from the detection of the sudden movement, but can also be the time from the detection of the sudden movement to the vehicle stopping. For example, the accident detection unit 125 determines that condition 3 is satisfied if the vehicle stops within a certain time (for example, within 10 seconds) after the detection of the sudden movement. Note that the 10 seconds period exemplified here is merely an example, and the setting can be changed as appropriate. Furthermore, the time can be measured by a general method, such as using time information or a timer in the accident detection device 100 or another in-vehicle device to measure the time from the detection of the sudden movement until the vehicle speed falls below a threshold.

[0069] As described above, when the detection result of the impact detection unit 124 determines that an abrupt movement has occurred, the accident detection unit 125 additionally determines whether or not all of the above-mentioned conditions 1 to 3 are satisfied. As a result, when all of the above-mentioned conditions 1 to 3 are satisfied, the accident detection unit 125 determines that contact occurred when the impact was detected, and determines that the driver has determined that this is an accident and has slowed down or stopped the vehicle, thereby properly detecting the impact that was erroneously detected as an abrupt movement as an accident.

[0070] (Explanation of scoring) Next, we will explain the process of detecting an accident by scoring the impact and sound performed by the accident detection unit 125. The accident detection unit 125 scores each of the impact and the sound inside the vehicle, and detects a vehicle accident based on the sum of the scores calculated using weights changed according to the determination result of the noise state.

[0071] For example, the accident detection unit 125 calculates a score according to the magnitude of the resultant vector or moving average value calculated by the impact detection unit 124, and the volume peak value of the analyzed audio data generated by the audio detection unit 121. The accident detection unit 125 then calculates a weight-adjusted score by applying a weight to the score that increases the weight of the impact and decreases the weight of the audio as the noise inside the vehicle increases. The accident detection unit 125 then calculates the total value of the weight-adjusted scores and detects a vehicle accident by comparing the total score with a preset threshold.

[0072] (Example) Next, the above-mentioned audio threshold change process and scoring process will be described with specific examples. Fig. 10 is a diagram showing an example of the processing contents of the change unit according to the first embodiment. Fig. 10 shows an example in which the volume threshold is changed depending on whether the noise state inside the vehicle determined by the determination unit 122 is "(1) noisy state" or "(2) quiet state."

[0073] As shown in FIG. 10(1), when the noise state inside the vehicle is noisy, the threshold value "350" for the contact volume judgment in the normal state (condition 1-(3)) before the change is exceeded at points other than the volume peak indicating the contact sound. In such a situation, the change unit 123 changes the threshold value for the contact volume judgment to "600" because the vehicle interior is in a noisy state. As a result, only the volume peak indicating the contact sound exceeds the threshold value "600", and the accident detection unit 125 can appropriately make the contact volume judgment related to condition 1-(3).

[0074] On the other hand, as shown in FIG. 10(2), when the noise state inside the vehicle is quiet, even the volume peaks indicating contact sound do not exceed the threshold value "350" of the contact sound volume determination in the normal state (condition 1-(3)) before the change. In such a situation, the change unit 123 changes the threshold value of the contact sound volume determination to "200" because the vehicle interior is quiet. As a result, only the volume peaks indicating contact sound exceed the threshold value "200," and the accident detection unit 125 can appropriately determine the contact sound volume determination related to condition 1-(3).

[0075] The values ​​of the threshold before and after the change are merely examples, and appropriate values ​​can be set depending on the vehicle model, occupant characteristics, etc. so that only volume peaks indicating contact sounds in the analyzed audio data generated by the audio detection unit 121 can be detected.

[0076] Next, a specific example of accident detection processing when using the score by the accident detection unit 125 will be described. Fig. 11 is a diagram showing an example of processing content of the accident detection unit when using the score according to embodiment 1. Fig. 11 shows a processing example in which the detection result of an impact that has occurred to the vehicle and the detection result of a sound inside the vehicle are each scored, and an accident is detected based on the total value of the scores weighted according to the noise level inside the vehicle.

[0077] First, the sound detection unit 121 analyzes sound data inside the vehicle during normal driving when no impact is detected, and outputs the analyzed sound data to the determination unit 122. Then, the determination unit 122 determines that the noise state inside the vehicle is a noisy state from the output analyzed sound data, and outputs the determination result to the modification unit 123. Next, based on the determination result that the state is a noisy state, the modification unit 123 increases the weight of the impact (1.0 → 1.5) and decreases the weight of the sound (1.0 → 0.5), and outputs the modification result to the accident detection unit 125.

[0078] In such a situation, when an impact occurs to the vehicle, the impact detection unit 124 outputs the impact detection result to the accident detection unit 125, and the audio detection unit 121 outputs the audio detection result before and after the impact detection to the accident detection unit 125.

[0079] The accident detection unit 125 calculates each score from the impact detection results and the audio detection results (FIG. 11(1)). For example, the accident detection unit 125 calculates "impact score: 8" and "audio score: 14". Next, the accident detection unit 125 applies the changed weight to each score (FIG. 11(2)). For example, the accident detection unit 125 multiplies "impact score: 8" by a weight of "1.5" to calculate "impact score: 12". Similarly, the accident detection unit 125 multiplies "audio score: 14" by a weight of "0.5" to calculate "audio score: 7".

[0080] Next, the accident detection unit 125 calculates the total score after weighting (FIG. 11(3)). For example, the accident detection unit 125 calculates the sum of the weighted "impact score: 12" and the weighted "audio score: 7" to obtain a "total score: 19". The accident detection unit 125 then compares the total score with a threshold value of "20 (as an example)" to determine whether or not an accident has occurred (FIG. 11(4)). For example, the accident detection unit 125 determines that the detected impact is not an accident because the "total score: 19" is less than the threshold value of "20".

[0081] Through the above-described series of processes, the accident detection device 100A applies weights that are changed according to the noise state inside the vehicle to the impact score and the audio score, and determines an accident from the total value of the adjusted scores. As a result, the accident detection device 100A can prevent erroneous determinations caused by the audio score being higher than it should be due to noise inside the vehicle, and can properly determine an accident.

[0082] (Processing flow of the accident detection device 100A) Next, an example of a processing procedure performed by the accident detection device 100A according to the first embodiment will be described with reference to Fig. 12. Fig. 12 is a flowchart showing an example of the flow of the accident detection processing according to the first embodiment. Note that the steps in the flowchart shown in Fig. 12 may be executed in a different order, and some processing may be added or omitted. Fig. 12 shows an example in which, when the accident detection device 100A detects a sudden movement, it determines whether or not an accident has occurred by determining the above-mentioned conditions 1 to 3.

[0083] First, the accident detection device 100A detects sounds inside the vehicle during normal driving (S101). Next, the accident detection device 100A determines the noise state inside the vehicle based on the detected sounds inside the vehicle (S102). Next, the accident detection device 100A changes the threshold or weight used to determine an accident according to the determination result of the noise state inside the vehicle (S103).

[0084] Then, the accident detection device 100A determines whether or not a sudden movement has been detected from the impact that occurred to the vehicle (S104). If a sudden movement has not been detected (S104; No), the accident detection device 100A returns to S101 and continues the process. On the other hand, if a sudden movement has been detected (S104; Yes), the accident detection device 100A determines whether or not a voice deviating from normal voices has been detected in the analyzed voice data including the time when the sudden movement was detected (Condition 1) (S105).

[0085] If a sound deviating from normal sounds is detected in the analyzed sound data (S105; Yes), the accident detection device 100A determines whether or not a sound deviating from normal sounds was detected near the time of the sudden movement detection (Condition 2) (S106).If a sound deviating from normal sounds was detected near the time of the sudden movement detection (S106; Yes), the accident detection device 100A determines whether or not the vehicle slowed down or stopped within 200 m of the sudden movement detection (Condition 3) (S107).

[0086] If the vehicle slows down or stops within 200 m of the sudden movement detection (S107; Yes), the accident detection device 100A determines that the detected sudden movement is an accident (S108) and ends the process. On the other hand, if no sound deviating from normal is detected in the analyzed sound data (S105; No), if no sound deviating from normal is detected near the time of the sudden movement detection (S106; No), or if the vehicle does not slow down or stop within 200 m of the sudden movement detection (S107; No), the accident detection device 100A determines that the detected sudden movement is not an accident (S109) and ends the process.

[0087] (effect) As described above, the accident detection device 100A according to the first embodiment includes the determination unit 122, the change unit 123, and the accident detection unit 125. The determination unit 122 determines the noise state inside the vehicle. The change unit 123 changes the threshold value of the sound used to detect an accident, or changes the weight of the sound when detecting an accident by combining an impact and sound, based on the determination result of the noise state inside the vehicle. The accident detection unit 125 detects a vehicle accident based on the impact occurring to the vehicle and the sound inside the vehicle.

[0088] As a result, the accident detection device 100A can appropriately detect accidents regardless of noise inside or outside the vehicle by adjusting the threshold or weight of the sound used for accident detection depending on the noise level inside the vehicle.

[0089] Furthermore, when the audio threshold is changed and the vehicle interior is determined to be quiet, the accident detection device 100A changes the audio threshold to a value smaller than the set value. As a result, even when the vehicle interior is quiet and noise is low, the accident detection device 100A can adjust the threshold so that only the contact sound generated when an impact is detected exceeds the threshold, thereby enabling appropriate accident detection.

[0090] Furthermore, when changing the audio threshold and when the noise inside the vehicle is determined to be noisy, the accident detection device 100A changes the audio threshold to a value greater than the set value. As a result, even when the noise inside the vehicle is loud, the accident detection device 100A can adjust the threshold so that only the contact sound generated when an impact is detected exceeds the threshold, thereby enabling appropriate detection of an accident.

[0091] Furthermore, when changing the weight of the voice, the accident detection device 100A reduces the weight of the voice as the noise inside the vehicle increases. As a result, the accident detection device 100A can reduce the contribution rate of the voice data in determining whether an accident has occurred, taking into account that the reliability of the voice data used to detect an accident decreases as the noise inside the vehicle increases.

[0092] Furthermore, the accident detection device 100A scores each of the impact and the sound inside the vehicle, and detects a vehicle accident based on the sum of the scores calculated using weights changed according to the noise state determination result. As a result, the accident detection device 100A can flexibly detect accidents by combining both detection results according to the noise state inside the vehicle, without setting thresholds for the impact detection result and the sound detection result.

[0093] [Embodiment 2] In the first embodiment, a technique for suppressing erroneous determinations during noisy or quiet conditions by dynamically changing the threshold for accident detection based on the noise inside the vehicle has been described. However, the method for suppressing erroneous determinations is not limited to this. For example, the accident detection device 100B according to the second embodiment can suppress erroneous determinations by correcting the value of audio data according to the loudness of the noise. Therefore, in the second embodiment, an accident detection device 100B that corrects the value of audio data according to the loudness of the noise will be described.

[0094] The overall configuration of the accident detection device 100B according to the second embodiment is similar to that of the accident detection device 100A according to the first embodiment, and therefore detailed description thereof will be omitted. Below, processing contents of the accident detection device 100B that differ from those of the accident detection device 100A will be described.

[0095] (Processing contents of the accident detection device 100B) The processing details of the accident detection device 100B according to the second embodiment will be described. The accident detection device 100B according to the second embodiment detects sounds inside the vehicle and calculates an average value of first sound values ​​detected during a first period during which the vehicle is normally traveling. The accident detection device 100B then detects a vehicle accident using sounds obtained by subtracting the average value of the first period from sounds acquired during a second period including the timing at which an impact of a predetermined value or greater that has occurred on the vehicle is detected.

[0096] Here, the processing contents of the accident detection device 100B will be described with a specific example. Fig. 13 is a diagram for explaining the processing contents of the accident detection device according to embodiment 2. Fig. 13 shows a processing example in which the value of audio data is corrected according to the volume of noise inside the vehicle.

[0097] For example, the accident detection device 100B calculates the average value of the noise as the average of the sum of the amplitudes of all frequency bands calculated at 100 ms intervals during normal vehicle operation. When the accident detection device 100B detects an impact on the vehicle as a sudden movement, it subtracts the average value of the noise from the sum of the amplitudes for the period including 7 seconds before and 1 second after the sudden movement was detected, and performs the determination process of conditions 1 to 3 described above using the audio obtained by the subtraction to detect an accident.

[0098] As a result, even if the impact that occurs to the vehicle is detected as a sudden movement that is less than an accident, the accident detection device 100B can properly detect a vehicle accident because only the volume peak at the time of detecting the sudden movement exceeds the adjusted volume threshold.

[0099] (Functional configuration of the accident detection device 100B) Next, an example of the functional configuration of the accident detection device 100B according to embodiment 2 will be described. Fig. 14 is a block diagram showing an example of the configuration of the accident detection device according to embodiment 2. As shown in Fig. 14, the accident detection device 100B differs from the accident detection device 100A according to embodiment 1 described above in that the detection processing unit 120 includes a calculation unit 126 instead of the determination unit 122 and the change unit.

[0100] The calculation unit 126 calculates an average value of the values ​​of the first voice detected during a first period during which the vehicle is normally traveling. The processing of the calculation unit 126 will be described below with a specific example.

[0101] For example, the calculation unit 126 calculates an average value of noise, which is the first sound, detected in a first period. Specifically, the calculation unit 126 calculates, for analyzed sound data generated at 100 ms intervals during normal vehicle travel, the average value of the sum of amplitudes of noise detected in the first period, which is defined as the first period from 7 seconds to 3 seconds immediately before a period during which it is estimated that no contact sound associated with an impact is detected. That is, the calculation unit 126 calculates the average value of the sum of amplitudes of noise, which is the first period during which the vehicle is traveling normally and no contact occurs, as noise other than contact sound.

[0102] Here, the first period mentioned above is not particularly limited as long as it is a period during which the vehicle is traveling normally without any impact being detected. For example, it may be the 5 seconds from 10 seconds to 5 seconds immediately before the time the impact is detected, or it may be the total period during which the total amplitude value is 300 or less in the immediately preceding 10 seconds.

[0103] Furthermore, the calculation unit 126 can also calculate the average value of the first audio value for each frequency band. Specifically, the calculation unit 126 calculates the average value of the amplitude values ​​for each frequency band during the first period for the amplitudes calculated for each frequency band by the audio detection unit 121. In this case, the audio detection unit 121 stores analysis data indicating the amplitude for each frequency band in the storage unit 110, rather than the total value of the amplitudes for all frequency bands.

[0104] The accident detection unit 125 detects a vehicle accident using audio obtained by subtracting the average value of the first audio from audio acquired within a second period including the timing at which an impact equal to or greater than a predetermined value is detected. For example, when an impact equal to or greater than a sudden movement is detected by the impact detection unit 124, the accident detection unit 125 subtracts the average value of the sum of the amplitudes of noise during the first period calculated by the calculation unit 126 from the sum of the amplitudes of audio detected within a second period, which is the period from 7 seconds before to 1 second after the detection of the impact. Then, the accident detection unit 125 uses the audio data after analysis of the subtracted second period to make judgments according to the above-mentioned conditions 1 to 3 and detects a vehicle accident.

[0105] For example, the accident detection unit 125 detects a vehicle accident using a sound obtained by subtracting the average value of the first sound from the average value of the second sound for each frequency band based on the average values ​​calculated for each frequency band. Explaining the above example, the accident detection unit 125 subtracts the average amplitude value calculated for each frequency band by the calculation unit 126 from the amplitude value of each frequency band of the sound for the period from 7 seconds before to 1 second after the time of detecting the impact. Then, the accident detection unit 125 detects a vehicle accident using the analyzed sound data for the subtracted second period.

[0106] The accident detection unit 125 can change the value to be subtracted for each frequency band, for example, to effectively subtract noise characteristic of a certain frequency band. For example, when a sound in a relatively low frequency band characteristic of noise other than contact sound is detected continuously at a constant cycle, the accident detection unit 125 can subtract the maximum value of the amplitude of the frequency band of the noise during the first period. As a result, the accident detection unit 125 can detect an accident using the sound from which the sound in the frequency band of the noise has been removed.

[0107] Here, the accident detection unit 125 can perform the above-described accident detection process when the impact detection unit 124 detects an impact indicating a sudden movement. Specifically, when the accident detection unit 125 detects an impact equal to or greater than a predetermined value that is less than the first threshold and equal to or greater than the second threshold, the accident detection unit 125 detects a vehicle accident using the sound obtained by subtracting the average value from the second sound.

[0108] For example, if the resultant vector calculated by the impact detection unit 124 is less than a first threshold value, which is a threshold value for determining whether an accident is minor or medium-sized, and is equal to or greater than a second threshold value for determining whether an accident is sudden, the accident detection unit 125 determines that the vehicle has experienced a sudden behavior, and performs the determination of conditions 1 to 3 described above to detect a vehicle accident.

[0109] The process of subtracting the average value of the first voice from the second voice is not limited to the process performed by the accident detection unit 125. For example, the voice detection unit 121 can refer to the average value of the first voice and store the voice obtained by subtracting the average value of the first voice from the second voice in the memory unit 110 as analyzed voice data.

[0110] Furthermore, as a method of removing noise from inside a vehicle, such as a loud car stereo, the accident detection device 100B can also obtain the sound of the car stereo or the like separately as input information and remove the sound of the car stereo or the like from the sound obtained by the sound detection unit 121 in a manner similar to noise canceling. Specifically, the accident detection device 100B can remove the sound of the car stereo or the like obtained separately when performing PCM (Pulse Code Modulation) on the sound obtained by the microphone 140, or can remove the sound of the car stereo or the like from the sound obtained by the sound detection unit 121 in a manner similar to noise canceling by performing FFT on the sound obtained by the microphone 140 and the sound of the car stereo or the like obtained separately in the processing of the FFT unit 121b.

[0111] (Example) Next, a specific example of the process of subtracting the first sound from the second sound performed by the accident detection device 100B will be described. Fig. 15 is a diagram showing a specific example of the process performed by the accident detection unit according to the second embodiment. Fig. 15 shows a process example in which the average value of the sum of the amplitudes of the sounds in a first period is subtracted from the sum of the amplitudes of the sounds in a second period.

[0112] 15, the accident detection device 100B defines the period from 3 seconds to 7 seconds during normal driving as a first period and calculates an average value 300 of the sum of the amplitudes of noise detected during this period.The accident detection device 100B then subtracts the average value 300 of the sum of the amplitudes of noise from the audio data after analysis of a second period that includes the period from 10 seconds when the sudden vehicle behavior is detected to 3 seconds to 11 seconds, which is used in the determination processes for Conditions 1 and 2 described above.

[0113] As a result, the accident detection device 100B can prevent noise other than the volume peak (around 9.7 seconds) indicating the contact sound detected around the time of impact detection from exceeding the threshold (350) for contact volume determination according to condition 1-(3), and can appropriately determine the volume peak equivalent to the contact volume.

[0114] (Processing flow of the accident detection device 100B) Next, an example of a processing procedure performed by the accident detection device 100B according to the second embodiment will be described with reference to Fig. 16. Fig. 16 is a flowchart showing an example of the flow of the accident detection processing according to the second embodiment. Note that the steps in the flowchart shown in Fig. 16 may be executed in a different order, and some processing may be added or omitted. Fig. 16 shows an example in which, when the accident detection device 100B detects a sudden movement, it determines whether or not an accident has occurred by determining the above-mentioned conditions 1 to 3.

[0115] First, the accident detection device 100B detects sounds inside the vehicle during normal driving (S201). Next, the accident detection device 100B calculates an average value of the sounds inside the vehicle detected during a first period during normal driving (S202).

[0116] The accident detection device 100B then determines whether or not a sudden movement has been detected from the impact that occurred on the vehicle (S203). If a sudden movement has not been detected (S203; No), the accident detection device 100B returns to S201 and continues processing. On the other hand, if a sudden movement has been detected (S203; Yes), the accident detection device 100B subtracts the average value of the audio in the first period from the audio in the second period that includes the time when the impact that occurred on the vehicle was detected (S204). The accident detection device 100B then determines whether or not an audio that deviates from normal audio has been detected in the analyzed audio data that includes the time when the sudden movement was detected (Condition 1) (S205).

[0117] If a sound deviating from normal sounds is detected in the analyzed sound data (S205; Yes), the accident detection device 100B determines whether or not a sound deviating from normal sounds was detected near the time of the sudden movement detection (Condition 2) (S206).If a sound deviating from normal sounds was detected near the time of the sudden movement detection (S206; Yes), the accident detection device 100B determines whether or not the vehicle slowed down or stopped within 200 m of the sudden movement detection (Condition 3) (S207).

[0118] If the vehicle slows down or stops within 200 m of the sudden movement detection (S207; Yes), the accident detection device 100B determines that the detected sudden movement is an accident (S208) and ends the process. On the other hand, if no sound deviating from normal is detected in the analyzed sound data (S205; No), if no sound deviating from normal is detected near the time of the sudden movement detection (S206; No), or if the vehicle does not slow down or stop within 200 m of the sudden movement detection (S207; No), the accident detection device 100B determines that the detected sudden movement is not an accident (S209) and ends the process.

[0119] (effect) As described above, the accident detection device 100B according to the second embodiment includes the sound detection unit 121, the impact detection unit 124, the calculation unit 126, and the accident detection unit 125. The sound detection unit 121 detects sound inside the vehicle. The impact detection unit 124 detects an impact occurring to the vehicle. The calculation unit 126 calculates an average value of a first sound value detected during a first period. The accident detection unit 125 detects a vehicle accident using sound obtained by subtracting the average value of the first period from sound acquired during a second period including the timing at which an impact equal to or greater than a predetermined value was detected.

[0120] As a result, the accident detection device 100B can appropriately detect accidents regardless of noise inside or outside the vehicle by correcting the sound value used for accident detection in accordance with the sound inside the vehicle during normal driving.

[0121] Furthermore, the accident detection device 100B detects audio during a first period during normal driving as noise, and subtracts the average value of the noise from the audio data used to detect accidents. As a result, the accident detection device 100B determines that audio detected during normal driving is noise that is not a contact sound that would be detected in an accident, and can omit the noise value from the audio data used in the process of determining whether it is a contact sound, so that it can properly detect accidents regardless of noise inside or outside the vehicle.

[0122] Furthermore, the accident detection device 100B detects a vehicle accident by calculating the average value of the amplitude of each frequency band for the noise audio data in the first time period, and subtracting the average value of the amplitude of the corresponding frequency band for the noise from the amplitude of each frequency band for the audio data in the second time period. As a result, the accident detection device 100B can detect an accident using audio data from which noise characteristic of a certain frequency band has been appropriately removed.

[0123] Furthermore, when the impact detected by the impact detection unit 124 is less than a first threshold for determining a small to medium accident and is equal to or greater than a second threshold for determining a sudden movement, the accident detection device 100B detects a vehicle accident using the sound obtained by subtracting the average value of the first sound from the second sound. As a result, when the accident detection device 100B determines a sudden movement from the impact that occurred on the vehicle, it can appropriately determine whether it is an accident or a sudden movement using the corrected sound data.

[0124] [Embodiment 3] Incidentally, the accident detection device 100A according to the first embodiment and the accident detection device 100B according to the second embodiment described above immediately perform STFT on the voice data acquired during normal driving and store the analyzed voice data in the memory unit 110, but the accident detection device 100 according to the present invention is not limited to this.

[0125] For example, the accident detection device 100 does not perform STFT when acquiring voice data during normal driving, but stores the pre-analysis voice data in the storage unit 110. Then, when the accident detection device 100 determines that an impact on the vehicle is a sudden movement, it can retroactively acquire the most recent voice data up to the point in time when the impact was detected, and generate analyzed voice data by performing STFT on the acquired voice data.

[0126] The following describes the processing details of the accident detection device 100 according to the third embodiment. Fig. 17 is a diagram illustrating the processing details of the accident detection device according to the third embodiment. Fig. 17 shows an example of a case where, regardless of the detection of a sudden movement, voice data is analyzed at the same time as it is acquired to generate analyzed voice data (Fig. 17(1)), and a case where, after a sudden movement is detected, the most recent stored voice data is retroactively acquired and analyzed to generate analyzed voice data (Fig. 17(2)).

[0127] 17(1), when detecting audio data immediately regardless of whether the detected audio data is a sudden movement, the accident detection device 100 calculates the sum of the amplitudes of the audio data acquired every 100 ms, generates analyzed audio data indicating the sum of the amplitudes of the audio data for a predetermined period of time, and stores the analyzed audio data in the storage unit 110. In this case, the accident detection device 100 repeatedly performs a series of processes from acquiring the audio data to storing the analyzed audio data until detecting a sudden movement.

[0128] When the accident detection device 100 detects a sudden movement, it uses the analyzed audio data stored in the memory unit 110, including the time when the impact was detected, to make a judgment based on the above-mentioned conditions 1 to 3, and detects a vehicle accident.

[0129] In contrast, as shown in FIG. 17(2), when analyzing audio data retroactively after detecting a sudden movement, the accident detection device 100 repeatedly stores the audio data acquired every 100 ms as is in the storage unit 110. When detecting a sudden movement, the accident detection device 100 retroactively acquires pre-analysis audio data stored in the storage unit 110, including the time when the sudden movement was detected, generates analyzed audio data indicating the sum of the amplitudes, and then performs the determinations under conditions 1 to 3 described above to detect a vehicle accident. As a result, the accident detection device 100 analyzes only limited audio data including the time when the sudden movement was detected, thereby reducing the processing load related to the analysis of the audio data and reducing power consumption.

[0130] Next, an example of the processing contents when the processing contents of the accident detection device 100 according to the embodiment 3 are applied to the accident detection device 100B according to the embodiment 2 will be described. For example, the accident detection device 100B repeatedly performs a process of storing the audio data acquired by the microphone 140 as is in the storage unit 110 until a sudden movement is detected.

[0131] Then, for example, when the accident detection device 100B detects a sudden movement due to an impact on the vehicle, it retrieves audio data stored in the memory unit 110 from a period of 7 seconds before to 1 second after the time the impact was detected, and then performs STFT on the retrieved audio data to generate analyzed audio data that indicates the sum of the amplitudes of all frequency bands for each time period.

[0132] The accident detection device 100B then calculates, for example, from the analyzed voice data, the average value of the sum of amplitudes for a first period, which is a period from 7 seconds before to 3 seconds before the time of detecting the impact.The accident detection device 100B then subtracts the calculated average of the sum of amplitudes from the analyzed voice data for a second period, which is a period from 7 seconds before to 1 second after the time of detecting the impact, and performs the determination of the above-mentioned conditions 1 to 3 to detect a vehicle accident.

[0133] (Processing flow of the accident detection device 100 according to the third embodiment) Next, an example of a processing procedure performed by the accident detection device 100 according to the third embodiment will be described with reference to Fig. 18. Fig. 18 is a flowchart showing an example of the flow of the accident detection processing according to the third embodiment. Note that the steps in the flowchart shown in Fig. 18 may be executed in a different order, and some processing may be added or omitted.

[0134] 18 shows an example in which the accident detection device 100 analyzes the voice data after detecting a sudden movement and determines whether the detected sudden movement is an accident. Note that it is assumed that the accident detection device 100 repeatedly performs the process of acquiring the voice data and storing it in the storage unit 110 before detecting the sudden movement.

[0135] First, the accident detection device 100 determines whether or not a sudden movement has been detected from an impact that has occurred to the vehicle (S301). If a sudden movement has not been detected (S301; No), the accident detection device 100 waits until a sudden movement is detected. On the other hand, if a sudden movement has been detected (S301; Yes), the accident detection device 100 retroactively acquires audio data from the storage unit 110, including the time when the sudden movement was detected (S302). Then, the accident detection device 100 analyzes the retroactively acquired audio data and calculates the sum of the amplitudes of all frequency bands for each time period (S303).

[0136] Then, the accident detection device 100 determines whether or not a sound deviating from normal sounds has been detected in the analyzed sound data including the time when the sudden behavior was detected (Condition 1) (S304). If a sound deviating from normal sounds has been detected in the analyzed sound data (S304; Yes), the accident detection device 100 determines whether or not a sound deviating from normal sounds has been detected near the time when the sudden behavior was detected (Condition 2) (S305). If a sound deviating from normal sounds has been detected near the time when the sudden behavior was detected (S305; Yes), the accident detection device 100 determines whether or not the vehicle slowed down or stopped within 200 m of the sudden behavior detection (Condition 3) (S306).

[0137] If the vehicle slows down or stops within 200 m of the sudden movement detection (S306; Yes), the accident detection device 100 determines that the detected sudden movement is an accident (S307) and ends the process. On the other hand, if no sound deviating from normal is detected in the analyzed sound data (S304; No), if no sound deviating from normal is detected near the time of the sudden movement detection (S305; No), or if the vehicle does not slow down or stop within 200 m of the sudden movement detection (S306; No), the accident detection device 100 determines that the detected sudden movement is not an accident (S308) and ends the process.

[0138] (effect) As described above, after detecting a sudden movement, the accident detection device 100 according to the third embodiment analyzes the voice data including the time when the sudden movement was detected. As a result, the accident detection device 100 omits the analysis related to the execution of STFT on the voice data during normal driving when no sudden movement was detected, and analyzes only the voice data used to determine an accident when a sudden movement was detected, thereby reducing the processing load related to the analysis of the voice data and reducing power consumption.

[0139] [Embodiment 4] Although the embodiments of the present invention have been described above, the present invention may be embodied in various different forms other than the above-described embodiments.

[0140] (Numbers, etc.) The numerical values, graphs, threshold values ​​(for example, arbitrary numerical values), etc. used in the above embodiment are merely examples and can be changed as desired.

[0141] Furthermore, if the noise state inside the vehicle is determined to be normal or the average value of the sum of the amplitudes to be subtracted is determined to be below a certain value, the threshold value or numerical value changed by the threshold value change process or audio data correction process described in the above embodiment can be determined to be the threshold value or numerical value before the change, and the noise inside and outside the vehicle can be determined to have been eliminated.

[0142] In the processing using audio data in the above embodiment, sound pressure in decibels (db) can also be used, for example.

[0143] [Hardware configuration] The accident detection device 100 according to the first, second, and third embodiments described above is realized, for example, by a computer 1000 configured as shown in FIG. 19. The accident detection device 100 will be described below as an example. FIG. 19 is a hardware configuration diagram showing an example of a computer that realizes the functions of the accident detection devices according to the first, second, and third embodiments. The computer 1000 has a CPU 1100, a RAM 1200, a ROM 1300, an HDD 1400, a communication interface (I / F) 1500, an input / output interface (I / F) 1600, and a media interface (I / F) 1700.

[0144] The CPU 1100 operates and controls each unit based on programs stored in the ROM 1300 or the HDD 1400. The ROM 1300 stores a boot program executed by the CPU 1100 when the computer 1000 starts up, programs that depend on the hardware of the computer 1000, and the like.

[0145] HDD 1400 stores programs executed by CPU 1100, data used by such programs, etc. Communication interface 1500 receives data from other devices via a predetermined communication network and sends it to CPU 1100, and transmits data generated by CPU 1100 to other devices via a predetermined communication network.

[0146] The CPU 1100 controls output devices such as a display and a printer, and input devices such as a keyboard and a mouse, via the input / output interface 1600. The CPU 1100 acquires data from the input devices via the input / output interface 1600. The CPU 1100 also outputs generated data to the output devices via the input / output interface 1600.

[0147] Media interface 1700 reads a program or data stored in recording medium 1800 and provides it to CPU 1100 via RAM 1200. CPU 1100 loads the program or data from recording medium 1800 onto RAM 1200 via media interface 1700 and executes the loaded program. Recording medium 1800 is, for example, an optical recording medium such as a DVD (Digital Versatile Disc) or a PD (Phase Change Rewritable Disc), a magneto-optical recording medium such as an MO (Magneto-Optical disk), a tape medium, a magnetic recording medium, or a semiconductor memory.

[0148] For example, when the computer 1000 functions as the accident detection device 100 according to the first, second, and third embodiments, the CPU 1100 of the computer 1000 executes programs loaded onto the RAM 1200 to realize the functions of the detection processing unit 120, the video encoder processing unit 150, the audio encoder processing unit 160, and the muxer unit 170. The CPU 1100 of the computer 1000 reads and executes these programs from the recording medium 1800, but as another example, the CPU 1100 may obtain these programs from another device via a predetermined communication network.

[0149] 〔others〕 Furthermore, among the processes described in each of the above embodiments, all or part of the processes described as being performed automatically can be performed manually, or all or part of the processes described as being performed manually can be performed automatically using known methods. In addition, the information including the processing procedures, specific names, various data, and parameters shown in the above documents and drawings can be changed as desired unless otherwise specified. For example, the various information shown in each drawing is not limited to the information shown in the drawings.

[0150] Furthermore, the components of each device shown in the figure are conceptual functional components and do not necessarily have to be physically configured as shown in the figure. In other words, the specific form of distribution and integration of each device is not limited to that shown in the figure, and all or part of them can be functionally or physically distributed and integrated in any unit depending on various loads, usage conditions, etc.

[0151] Furthermore, the above-described embodiments can be combined as appropriate within the scope of not causing any contradiction in the processing content.

[0152] Although some of the embodiments of the present application have been described in detail above with reference to the drawings, these are merely examples, and the present invention can be implemented in other forms that include the embodiments described in the Disclosure of the Invention section and that have undergone various modifications and improvements based on the knowledge of those skilled in the art.

[0153] Furthermore, the above-mentioned "section, module, unit" can be read as "means" or "circuit," etc. For example, a receiving section can be read as receiving means or receiving circuit. [Explanation of symbols]

[0154] 100, 100A, 100B Accident detection device 110 Storage section 120 Detection processing unit 121 Voice detection unit 121a Window hanging part 121b FFT section 122 Judgment section 123 Changes 124 Impact detection unit 125 Accident detection unit 126 Calculation Unit 130 Camera 140 Mike 150 Video encoder processing unit 160 Audio Encoder Processing Unit 170 Muxer section 180 Accelerometer 200 External server device

Claims

1. a voice detection unit that detects voices inside the vehicle; an impact detection unit that detects an impact occurring to the vehicle; a calculation unit that calculates an average value of values ​​of the first audio detected in a first period; an accident detection unit that detects an accident involving the vehicle using the sound obtained by subtracting the average value for the first period from the sound acquired within a second period including a timing at which the impact equal to or greater than a predetermined value is detected; An accident detection device comprising:

2. The voice detection unit Detecting noise contained in the voice inside the vehicle; The calculation unit calculating an average value of the noise detected during a first period of time; The accident detection unit detecting an accident of the vehicle using audio obtained by subtracting the average value of the noise from audio for detecting an accident acquired within a second period including the time point at which the impact equal to or greater than a predetermined value is detected; 2. The accident detection device according to claim 1, wherein:

3. The calculation unit Calculating an average value of the first audio signal for each frequency band; The accident detection unit detecting an accident of the vehicle using a sound obtained by subtracting the average value calculated for each of the frequency bands from the sound acquired within the second period for each of the frequency bands; 2. The accident detection device according to claim 1.

4. The accident detection unit When an impact equal to or greater than the predetermined value that is less than the first threshold and equal to or greater than the second threshold is detected, an accident of the vehicle is detected using the sound obtained by subtracting the average value from the sound acquired within the second period.

2. The accident detection device according to claim 1.

5. a storage unit that stores the sound detected by the sound detection unit; The voice detection unit When the impact equal to or greater than the predetermined value is detected, a short-time Fourier transform is performed on the audio for a predetermined time prior to the time when the impact was detected, to calculate the amplitude of the audio for each predetermined time period; The calculation unit calculating an average value of the amplitude of the audio during the first period; The accident detection unit detecting an accident of the vehicle using the amplitude of the sound obtained by subtracting the average value of the amplitude during the first period from the amplitude during the second period of the sound; 2. The accident detection device according to claim 1.

6. 1. An accident detection method performed by an accident detection device, comprising: a voice detection step of detecting voices inside the vehicle; an impact detection step of detecting an impact occurring to the vehicle; calculating an average value of the first audio detected during the first time period; an accident detection step of detecting an accident involving the vehicle using the sound obtained by subtracting the average value for the first period from the sound acquired within a second period including a timing at which the impact equal to or greater than a predetermined value is detected; 10. An accident detection method comprising:

7. a voice detection step for detecting voices inside the vehicle; an impact detection step for detecting an impact occurring to the vehicle; calculating an average value of the first audio signal detected during the first time period; an accident detection step of detecting an accident involving the vehicle using the sound obtained by subtracting the average value for the first period from the sound acquired within a second period including a timing at which the impact equal to or greater than a predetermined value is detected; An accident detection program characterized by causing a computer to execute the above.

Images