JP2019014392A - Traveling recording apparatus for vehicle, and browsing device - Google Patents

Abstract

To provide a traveling recording apparatus for recording vehicle traveling situations, which can record information capable of facilitating a grasp of situations of unsafe incidents by detecting the unsafe incidents during a vehicle traveling motion.SOLUTION: A traveling recording apparatus includes: a travel information recording part that acquires travel information including an image of the periphery of a vehicle during vehicle traveling and records it; a request part that makes a request of explanation of situations to an occupant when a predetermined change of state occurs during the vehicle traveling; a speech recognition part that recognizes the content of utterance, given by the occupant, for the request of the request part; and an unsafe-incident determination part that determines whether or not the travel information recorded in the travel information recording part on the basis of the occupant's content of utterance recognized by the speech recognition part is an unsafe incident scene.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a travel recording device and a browsing device that record a travel state of a vehicle.

  As a vehicle travel recording device, as described in Patent Document 1, when the acceleration during vehicle travel exceeds a threshold, it is determined whether an accident has occurred from surrounding sounds. 2. Description of the Related Art A drive recorder configured to record surrounding images captured before and after is known.

JP 2006-199204 A

  The above drive recorder records the surrounding video at the time of the accident so that the situation at the time of the accident can be grasped from the recorded video after the accident occurs. There was a problem that it was difficult to detect.

  In other words, a near-miss is a situation one step before an accident that is not strange even if an accident occurs. And if such a near-miss is detected and the image | video and sound at that time can be recorded as a near-miss scene, the driving | operation manager of a vehicle can implement the guidance of the safe driving with respect to a driver based on the recorded information.

  However, in near-miss scenes, even if the acceleration exceeds the threshold, sounds similar to those at the time of the accident such as collision sound, brake sound, horn sound, etc. may not be generated, so video is recorded based on surrounding sounds. However, the drive recorder cannot record a near-miss scene where no sound similar to that at the time of the accident occurs.

  Even if the drive recorder can record a near-miss scene that did not result in an accident, it is difficult for the operation manager to identify the cause of the near-miss from the recorded video and sound. There is also the problem that it is difficult to do.

  In one aspect of the present disclosure, it is desirable that a traveling recording apparatus that records a traveling state of a vehicle can detect a near-miss while the vehicle is traveling and record information that can easily grasp the near-miss state.

In the vehicle travel recording apparatus according to one aspect of the present disclosure, the travel information recording unit (12, 14) acquires and records travel information including images around the vehicle while the vehicle is traveling.
When a predetermined state change occurs while the vehicle is traveling, the request unit (22) requests a situation explanation from the occupant, and the voice recognition unit (24) recognizes the utterance content uttered by the occupant in response to the request. To do.

  Then, the near-miss determining unit (32) determines whether the travel information recorded in the travel information recording unit is a near-miss scene based on the utterance contents of the occupant recognized by the voice recognition unit.

  As described above, according to the traveling recording device of the present disclosure, when a predetermined state change occurs during traveling of the vehicle, the utterance content of the situation explanation is acquired from the occupant, and the utterance content is recorded in the traveling information recording unit It can be detected that the running information is a near-miss scene.

  Therefore, according to the traveling recording apparatus of the present disclosure, the near-miss detection accuracy can be improved and the near-scene recorded in the traveling information recording unit can be accurately identified. For this reason, if the travel information of the identified near-miss scene is recorded in the nonvolatile recording medium (40), the cause of the near-miss is identified based on the recorded travel information after the vehicle travels. Can provide guidance.

  In addition, every time a near-miss is detected, if the utterance content of the occupant used to determine the near-hat is recorded in a non-volatile recording medium in addition to the driving information of the near-miss scene, the driver can A near-miss scene that is likely to occur can be identified. Therefore, in this case, driving guidance for the driver can be performed more appropriately.

  Further, when browsing the near-miss scene from the travel information recorded on the non-volatile recording medium using the browsing device (50), if the utterance content corresponding to the scene is output, the near-hat from the utterance content. The scene can be easily grasped.

  Note that the reference numerals in parentheses described in this column and in the claims indicate the correspondence with the specific means described in the embodiment described later as one aspect, and the technical scope of the present disclosure It is not limited.

It is a block diagram showing the structure of the travel recording apparatus and browsing apparatus of embodiment. It is a flowchart showing operation | movement of a driving | running | working information recording part. It is a flowchart showing operation | movement of the near-miss primary determination part. It is a flowchart showing operation | movement of a driving scene primary recording part. It is a flowchart showing operation | movement of a condition description request part. It is a flowchart showing the request sentence production | generation process shown in FIG. It is a flowchart showing operation | movement of a speech recognition part. It is a flowchart showing operation | movement of a near-miss secondary determination part. It is a flowchart showing the near-miss secondary determination score calculation process shown in FIG. It is a flowchart showing operation | movement of a driving | running | working scene secondary recording part. It is a flowchart showing operation | movement of a driving scene extraction part and a driving scene display part.

Hereinafter, embodiments of the present disclosure will be described with reference to the drawings.
The travel recording apparatus 10 according to the present embodiment is mounted on a vehicle such as an automobile, detects a near-miss generated while the vehicle travels, and records the image and vehicle state at that time as travel information on a non-volatile recording medium. belongs to.

The travel information thus recorded is used by a vehicle operation manager to browse using the browsing device 50 after the vehicle travels, and to give instructions on safe driving to the driver.
As shown in FIG. 1, the travel recording apparatus 10 includes a travel information recording unit 12, a near-miss primary determination unit 16, a travel scene primary recording unit 18, a situation explanation request unit 22, a voice recognition unit 24, and a near-miss secondary determination unit 32. And a travel scene secondary recording unit 38 is provided.

The traveling recording apparatus 10 is configured by an electronic control unit (ECU) centered on a microcomputer including a CPU, a ROM, a RAM, and the like. And each said part is implement | achieved when CPU runs various programs along the flowchart illustrated in FIGS.

  The travel information recording unit 12 acquires the behavior of the vehicle, the video around the vehicle, the position of the vehicle, the sound inside and outside the vehicle, the attributes of surrounding objects, and the like as travel information from other in-vehicle devices while driving the vehicle. A driver ID, which is the identification information of the driver, is assigned and recorded in the travel information DB 14. DB represents a database.

  Here, the behavior of the vehicle is the vehicle speed and acceleration, the amount of operation of the accelerator, brake, steering, etc. by the driver, and the travel information recording unit 12 controls the driving / braking system of the vehicle based on such information. Obtain from the control device.

  In addition, the images around the vehicle are images in front and rear in the vehicle traveling direction, and the left and right sides, and the traveling information recording unit 12 acquires these images from one or more cameras mounted on the vehicle. . Further, the travel information recording unit 12 acquires the position of the vehicle from a vehicle-mounted device that can detect the position, such as a GPS receiver, and acquires sound inside and outside the vehicle through a microphone mounted on the vehicle.

  The attributes of surrounding objects are the types, distances, relative speeds, etc. of objects around the vehicle that can be estimated by existing object recognition methods from information around the vehicle and information from the ranging sensor. The unit 12 acquires these pieces of information from a radar device, a navigation device, or the like.

  The travel information DB 14 is configured by a volatile recording medium. Then, the travel information recording unit 12 repeatedly executes the travel information recording process shown in FIG. 2 at a predetermined period during vehicle travel, thereby acquiring travel information from other in-vehicle devices and recording it in the travel information DB 14.

  As shown in FIG. 2, the travel information recording process measures the various travel information described above at S110, and at S120, the measured travel information is updated from the old information to the recording medium constituting the travel information DB 14. It is executed by the procedure of overwriting in order.

As a result, the traveling information for the past predetermined period is recorded in the traveling information DB 14 from the latest one.
Next, the near-miss primary determination unit 16 primarily determines whether or not a near-miss has occurred based on the travel information recorded in the travel information DB 14, and repeats the near-miss primary determination process shown in FIG. 3 at a predetermined cycle. Run.

  As shown in FIG. 3, in the near-miss primary determination process, the latest acceleration recorded as one of the travel information in the travel information DB is acquired in S210, and in the subsequent S220, the acceleration is a preset threshold value. Judge whether or not

  If it is determined in S220 that the acceleration exceeds the threshold value, the process proceeds to S230, the primary recording flag and the situation explanation request flag are set to “TRUE”, and the near-miss primary determination process ends.

If it is not determined in S220 that the acceleration exceeds the threshold value, the near-miss primary determination process is terminated as it is.
In addition, since the determination process of S220 is for detecting a near-miss from a change in the state of the vehicle, it is determined whether or not the absolute value of the acceleration including the deceleration of the vehicle has exceeded a threshold value. Further, in S220, it is only necessary to determine a near-miss from the vehicle state change that occurs when a near-miss occurs. For example, the near-miss may be determined using the vehicle speed or the steering angle of the steering. Moreover, the threshold value used for near-miss determination may be set in advance as a fixed value, or may be arbitrarily set by an operation manager or the like.

  Next, the primary recording flag set to “TRUE” in S230 is used in the traveling scene primary recording unit 18 to determine whether or not to record the traveling information in the traveling scene primary recording DB 20.

  That is, the traveling scene primary recording unit 18 repeatedly executes the traveling scene primary recording process shown in FIG. 4 at a predetermined period, so that the traveling information recorded in the traveling information DB 14 when the primary recording flag is “TRUE”. Is recorded in the travel scene primary record DB.

  As shown in FIG. 4, in the traveling scene primary recording process, first, in S310, it is determined whether or not the primary recording flag is “TRUE”. If the primary recording flag is “TRUE”, the process proceeds to S320. If the primary recording flag is not “TRUE”, the traveling scene primary recording process is terminated.

  In S320, travel information for a predetermined period before and after the determination time when the near-miss primary determination unit 16 determines that a near-miss has occurred is acquired from the travel information DB 14, and the process proceeds to S330. In S320, for example, the travel information from the determination time to 5 seconds before and the travel information of the longer period until 5 seconds have elapsed from the determination time or until later-described voice recognition ends are acquired.

  In S330, the acquired travel information for a predetermined time is recorded in the travel scene primary record DB 20 as a travel scene representing a near-miss situation, and in S340, the primary record flag is set to “FALSE”. Then, the running scene primary recording process is terminated.

  The traveling scene primary recording DB 20 is configured by a volatile recording medium, like the traveling information DB 14, and a recording area capable of recording at least one scene of a near-miss situation is secured.

  For this reason, when a travel scene is recorded in the travel scene primary record DB 20 in S330, if a past travel scene is recorded in the travel scene primary record DB 20, the recorded travel scene is recorded in S320. The acquired latest driving scene is overwritten.

  Next, the situation explanation request unit 22 is for urging the driver to explain the situation of the driving scene when the near-miss primary judgment unit 16 determines that the near-miss is detected. The situation explanation request process shown in FIG. Are repeatedly executed in a predetermined cycle.

  As shown in FIG. 5, in the situation explanation request process, in S410, it is determined whether or not the situation explanation request flag is “TRUE”. If the situation explanation request flag is “TRUE”, the process proceeds to S420. Otherwise, the situation explanation request process is terminated.

In S420, the speech recognition flag is set to “TRUE”, the process proceeds to S430, and the request sentence generation process is executed.
This request sentence generation process is a process for generating a request sentence for prompting the driver to explain the situation, based on the traveling information when the near-miss primary determination unit 16 determines a near-miss, for example, the procedure shown in FIG. Is executed.

That is, in the request sentence generation process, in S432, it is determined whether an object exists around the vehicle based on the travel information. If an object exists around the vehicle, the process proceeds to S434, and the vehicle Get the attribute of the closest object. That is, in S434, it is estimated that the object closest to the vehicle is the causative object that causes the near-miss, and the attribute of the object is acquired.

  In the subsequent S436, based on the attribute of the object acquired in S434, a request sentence for asking the driver whether or not it is likely to collide with the object is determined, and the request sentence generation process ends.

  If it is determined in S432 that there is no object around the vehicle, the process proceeds to S438 to determine a request sentence for asking the driver whether or not the object is about to collide with an object outside the measurement range. The request statement generation process is terminated.

  For example, in S438, since the object is not recognized, a request sentence for asking the driver about the situation is generated as "What happened?" In S436, since the attribute of the object closest to the vehicle is recognized, based on the attribute, for example, "There was a person in the front, but did you seem to hit that person?" Generate a request statement.

  By generating a request sentence in this way, for example, a request sentence for actively explaining the cause of the near-miss to the driver can be generated, such as "No. A person came out from the side." It becomes like this.

  In other words, when it is likely to collide with a person who has jumped out from the side of the vehicle, and the near-miss primary determination unit 16 determines that it is a near-miss, the driver is uniformly asked "What has happened?" , "It seems to have hit a person" is considered to be answered.

  However, in such an answer, if the person who jumped out from the side of the vehicle does not appear in the image while the vehicle is running when viewed after the vehicle is running, the operation manager will be shown in the image while the vehicle is running. It is judged that the person is likely to collide with the other person, and the driver will overlook that it is difficult to notice the pedestrian jumping out from the side.

  However, if the cause of the near-miss is estimated and a request sentence is generated based on the estimation result as in this embodiment, the cause of the near-miss is more accurately identified at the time of browsing, and driving guidance is properly performed. become able to.

  In S434, the object closest to the vehicle is estimated as the object causing the near-miss, but for example, the object having the fastest approach speed is estimated as the object causing the near-miss, and the attribute of the object is determined. You may make it acquire. Further, the type of object, distance, relative speed, etc. may be comprehensively determined to estimate the object causing the near-miss.

  Next, when a request sentence is generated in S430, the process proceeds to S440, where the generated request sentence is voice-synthesized and the voice is output from the speaker, thereby requesting the driver to explain the situation. .

In S450, since the situation explanation request is completed, the situation explanation request flag is set to "FALSE", and the situation explanation request process is terminated.
The request text may be output by, for example, prompting the driver to explain the situation by displaying the request text on the windshield of the vehicle or the like, and prompting the situation explanation by both voice output and display. You may do it.

  Next, the voice recognition unit 24 recognizes the voice uttered by the driver in response to the request from the situation explanation request unit 22, and outputs the utterance word as the recognition result as text data representing the utterance content. For example, the voice recognition process shown in FIG. 7 is repeatedly executed at a predetermined cycle.

  As shown in FIG. 7, in the speech recognition process, first, in S510, it is determined whether or not the speech recognition flag is “TRUE”. If the speech recognition flag is “TRUE”, the process proceeds to S520. Otherwise, the speech recognition process is terminated.

  In S520, it is determined whether or not a predetermined time: t seconds have elapsed since the voice recognition flag is set to “TRUE” and voice recognition in S530 and S540 is started. If the predetermined time: t seconds or longer has not elapsed, the process proceeds to S530, and if the predetermined time: t seconds or longer has elapsed, the process proceeds to S550.

  In S530, the sound inside and outside the vehicle input from the microphone is measured, and in S540, the most likely word is output as a speech recognition result, and the process proceeds to S520. That is, in S530 and S540, speech recognition is performed using the acoustic model 26 and the language model 28 shown in FIG.

  The acoustic model 26 and the language model 28 are, for example, hidden Markov models and are recorded in advance in a nonvolatile memory. In addition, the dictionary included in the language model may be a general spoken language, for example, composed of near-miss words such as “intersection” and “bicycle” and negative words such as “nothing” or “careful” Only the secondary determination word may be used. If the dictionary of language models is limited to near-miss related words and negative words, words necessary for near-miss secondary determination described later can be recognized more accurately.

  On the other hand, in S550, since the predetermined time: t seconds or more have elapsed after the start of the speech recognition, the speech recognition flag is set to “FALSE” and the speech recognition is terminated. In the subsequent S560, the near-miss secondary determination flag is set to “TRUE”, and the speech recognition process ends.

  Next, the near-miss secondary determination unit 32 determines whether or not a near-miss actually occurred from the driver's utterance word recognized by the voice recognition unit 24. For example, FIG. The near-miss secondary determination process shown is repeatedly executed at a predetermined cycle.

  As shown in FIG. 8, in the near-miss secondary determination process, in S610, it is determined whether the near-miss secondary determination flag is “TRUE”. If the near-miss secondary determination flag is “TRUE”, the process proceeds to S620. If not, the near-miss secondary determination process is terminated.

In S620, a near-miss secondary determination score calculation process for calculating a score used for the near-miss secondary determination in the procedure shown in FIG. 9 is executed.
As shown in FIG. 9, in the near-miss secondary determination score calculation process, in S <b> 622, the acceleration value used for the near-miss primary determination by the near-miss primary determination unit 16 is acquired as a score 1.

  In subsequent S624, the ratio of near-miss related words and negative words included in the utterance words recognized by speech recognition unit 24 for a predetermined time: t seconds is obtained, and the difference between the ratios of these words is scored. 2 is calculated.

Finally, in S626, the weighted sum of score 1 and score 2 is obtained by weighted addition of score 1 acquired in S622 and score 2 calculated in S624, and this is set as the secondary determination score. .

  The near-miss related words and the negative words are set in advance, and are recorded in the nonvolatile memory as the near-hat related word dictionary 34 and the negative word dictionary 36 as in the acoustic model 26 and the language model 28.

  As described above, when the secondary determination score is calculated in the near-miss secondary determination score calculation process, the process proceeds to S630, and it is determined whether or not the calculated secondary determination score is equal to or greater than a preset threshold value. To do.

  That is, in S630, by comparing the secondary determination score and the threshold value, it is determined whether the ratio of near miss related words is higher than the ratio of negative words by a predetermined ratio or more in the speech-recognized utterance words. It is determined whether or not a near-miss has occurred.

  Further, since the near-miss primary determination unit 16 also adds the weight of the acceleration used for the primary determination to the secondary determination score, the near-miss can be accurately determined in S630.

  That is, for example, if the acceleration exceeds the threshold value and the driver does not feel like a near-miss, the greater the acceleration, the easier it is to determine the near-miss in S630, thus increasing the near-miss determination accuracy. it can.

  If it is determined in S630 that the near-miss secondary determination score is equal to or greater than the threshold, the process proceeds to S640, and if not, the process proceeds to S650. In S640, the near-miss secondary recording flag is set to “TRUE”, and the process proceeds to S650.

In S650, the near-miss secondary determination flag is set to “FALSE”, and the near-miss secondary determination process ends.
Next, the traveling scene secondary recording unit 38 acquires the determined traveling scene from the traveling scene primary recording DB 20 when the near-miss secondary determining unit 32 determines that a near-miss has occurred, and the traveling scene secondary recording unit 38 obtains the determined traveling scene. This is for recording in the recording DB 40.

That is, the traveling scene secondary recording unit 38 functions as a scene recording unit of the present disclosure by repeatedly executing the traveling scene secondary recording process illustrated in FIG. 10 at a predetermined period.
As shown in FIG. 10, in the traveling scene secondary recording process, first, in S710, it is determined whether or not the secondary recording flag is “TRUE”. If the secondary recording flag is “TRUE”, the process proceeds to S720. If the secondary recording flag is not “TRUE”, the traveling scene secondary recording process is terminated.

  In S720, the driving scene in which the near-miss secondary determination unit 32 determines that a near-miss occurred is acquired from the driving scene primary record DB 20, and the process proceeds to S730. In S730, the voice recognition result used to determine the near-miss is determined from the near-miss secondary determination unit 32, and the acquired voice recognition result and the driving scene acquired in S720 are stored in the driving scene secondary recording DB 40. Record.

  The traveling scene secondary recording DB 40 is configured by a non-volatile recording medium so that a traveling scene of a near-miss can be maintained even if power supply to the traveling recording apparatus 10 is interrupted after traveling the vehicle.

In S730, when the traveling scene is recorded in the traveling scene secondary recording DB 40 together with the voice recognition result, the process proceeds to S740, where the secondary recording flag is set to “FALSE”, and the traveling scene secondary recording process is performed. Exit.

  As described above, in the vehicle travel recording apparatus 10 of the present embodiment, during travel of the vehicle, the travel information recording unit 12 acquires travel information including images around the vehicle and records it in the travel information DB 14.

  When the near-miss primary determination unit 16 determines a near-miss while the vehicle is running, the situation explanation request unit 22 requests a situation explanation from the driver or the like, and the voice recognition unit 24 speaks to the request. Recognize utterance content.

  Then, the near-miss secondary determination unit 32 secondarily determines the near-miss based on the content of the utterance, and when the near-miss secondary determination unit 32 also determines the near-miss, the travel information at that time is used as a near-miss travel scene. Recorded in the scene secondary recording DB 40.

  For this reason, according to the traveling recording apparatus 10 of the present embodiment, the near-miss can be determined in two stages based on the acceleration of the vehicle and the utterance content of the occupant, and the occurrence of the near-miss can be accurately detected.

  The travel scene recorded in the travel scene secondary record DB 40 is composed of travel information such as vehicle behavior, vehicle surroundings, vehicle position, sound inside and outside the vehicle, attributes of surrounding objects, driver ID, and the like. ing. In addition to the traveling scene, text data that is a recognition result by the voice recognition unit 24 is also recorded in the traveling scene secondary recording DB 40.

  Therefore, according to the travel recording apparatus 10 of the present embodiment, after the vehicle travels, the travel scene recorded in the travel scene secondary recording DB 40 is browsed via the viewing apparatus 50 to identify the cause of the near-miss. , Driving guidance of the driver can be performed.

  Further, when browsing a driving scene, text data that is a voice recognition result for each driving scene of the near hat can be used to classify the cause of the near hat and easily identify the near hat scene that is likely to occur by the driver. Therefore, if the vehicle travel recording apparatus 10 of the present embodiment is used, the driver's driving guidance can be performed more appropriately.

Next, the browsing device 50 used for the vehicle operation manager to browse the travel scene recorded in the travel scene secondary record DB 40 will be described.
As shown in FIG. 1, the browsing device 50 displays a travel scene extraction unit 52 that extracts a travel scene from the travel scene secondary record DB 40 according to a query specified by the operation manager, and the extracted travel scene on the display. And a traveling scene display section 54 for performing the operation.

The travel scene extraction unit 52 and the travel scene display unit 54 are realized by a computer constituting the browsing device 50 executing the travel scene extraction / display process shown in FIG.
That is, in the travel scene extraction / display process, in S810, a process as the travel scene extraction unit 52 that extracts a travel scene that matches the query specified by the operation manager from the travel scene secondary record DB 40 is executed.

Examples of the query specified by the operation manager include a driver ID, a spoken word, acceleration, the presence / absence of a surrounding object, a distance from the surrounding object, and the type of the surrounding object.
Next, in S820, it is determined whether the display form of the traveling scene is a list display or an individual display. This display mode determination is performed in accordance with a display command from the operation manager. When a list is displayed, the process proceeds to S830, and when individually displayed, the process proceeds to S840.

  In S830, the query used to extract the driving scene in S810, the number of extracted driving scenes, the utterance words assigned to the extracted driving scene and the frequency of occurrence thereof, etc. are displayed in a list on the display.

  For this reason, for example, when a driving scene of a specific driver is extracted using the driver ID in S810, for example, “person” 60%, “bicycle” 20%, “intersection” are displayed by list display. 80%, etc., the utterance word and its occurrence frequency are displayed.

  In this case, the operation manager can determine from the list display screen that the driver specified by the driver ID is “easy to collide with a person at the intersection”, and can appropriately perform the driving instruction of the driver. It becomes like this.

In S840, one travel scene to be displayed is selected from the travel scenes extracted in S810, and the process proceeds to S850.
In S850, the driver ID, the vehicle behavior, the video, the position, the peripheral object attribute, the voice, and the spoken word are displayed on the display according to various information included in the selected travel scene. Note that audio is output from a speaker.

  As a result, the operation manager can quickly grasp the outline of the driving scene from the utterance word displayed in the text on the display, and the cause of the near-miss can be obtained without looking at the video of the driving scene to the end. It becomes possible to specify.

As mentioned above, although the form for implementing this invention was demonstrated, this invention is not limited to the above-mentioned embodiment, It can implement in various deformation | transformation.
For example, in the above embodiment, the voice recognition unit 24 has been described as recognizing the utterance language uttered by the occupant in the processes of S530 and S540. You may make it recognize accident sound.

In this way, even if the driver cannot explain the situation or lies that it is not a near-miss, it can detect a near-miss.
The voice recognition by the voice recognition unit 24 may be started before the situation explanation request unit 22 ends the request for situation explanation, for example, immediately after the near-miss primary judgment unit 16 determines the near-miss. .

In this way, even if the driver starts speaking without waiting for a request, the driver can recognize the content of the speaking.
In addition, although the voice recognition by the voice recognition unit 24 has been described as being performed for t seconds after a request for situation explanation, this time may be a fixed time such as 10 seconds or may be set according to a silent period. May be.

  That is, for example, if the silent state continues for 3 seconds or more, the voice recognition may be completed. In this way, even if the explanation by the driver is long, the utterance content can be recognized until the end of the explanation.

In this case, the silent section may be determined as the silent section if the amplitude of the audio signal is less than a certain threshold or the number of zero crossings is less than the threshold.
Next, in the above embodiment, the travel recording apparatus 10 is configured by an electronic control unit configured by a microcomputer, and each unit illustrated in FIG. 1 is realized by the CPU executing a program. As explained.

  However, each part of the travel recording apparatus 10 may be realized partially or entirely using hardware that combines a logic circuit, an analog circuit, and the like. The same applies to the travel scene extraction unit 52 and the travel scene display unit 54 of the browsing device 50.

  In addition, a plurality of functions of one constituent element in the above embodiment may be realized by a plurality of constituent elements, or a single function of one constituent element may be realized by a plurality of constituent elements. Further, a plurality of functions possessed by a plurality of constituent elements may be realized by one constituent element, or one function realized by a plurality of constituent elements may be realized by one constituent element. Moreover, you may abbreviate | omit a part of structure of the said embodiment. In addition, at least a part of the configuration of the above embodiment may be added to or replaced with the configuration of the other embodiment. In addition, all the aspects included in the technical idea specified only by the wording described in the claim are embodiment of this invention.

  DESCRIPTION OF SYMBOLS 10 ... Travel recording device, 12 ... Travel information recording part, 14 ... Travel information DB, 16 ... Near-miss primary determination part, 22 ... Situation description request part, 24 ... Voice recognition part, 32 ... Near-miss secondary determination part, 38 ... Travel Secondary scene recording unit, 40 ... travel scene secondary record DB, 50 ... browsing device, 52 ... travel scene extraction unit, 54 ... travel scene display unit.

Claims (13)

A travel information recording unit (12, 14) for acquiring and recording travel information including images around the vehicle during vehicle travel;
A request section (22) for requesting a situation explanation to an occupant when a predetermined state change occurs while the vehicle is running;
A voice recognition unit (24) for recognizing the utterance content spoken by the occupant in response to the request of the request unit;
A near-miss determining unit (32) for determining whether the traveling information recorded in the traveling information recording unit is a near-miss scene based on the utterance content of the occupant recognized by the voice recognition unit;
A vehicle travel recording apparatus comprising:   The vehicle travel recording apparatus according to claim 1, further comprising a scene recording unit (38) for recording the travel information determined to be a near-miss scene by the near-miss determination unit on a non-volatile recording medium (40). .   In addition to the travel information recorded in the travel information recording unit, the scene recording unit stores the utterance content used by the near-miss determination unit to determine that the near-scene is the near-scene on the nonvolatile recording medium. The vehicle travel recording apparatus according to claim 2, configured to record.   When recording the utterance content on the nonvolatile recording medium, the scene recording unit is configured to record text data of the utterance content recognized by the voice recognition unit on the nonvolatile recording medium. The vehicle travel recording apparatus according to claim 3. The travel information recording unit acquires and records the status of an object existing around the vehicle as one of the travel information,
The request unit determines the content of the request based on the state of the object recorded in the travel information recording unit.
The vehicle travel recording apparatus according to claim 1, configured as described above.   The travel recording apparatus according to claim 5, wherein the request unit is configured to estimate a cause object of a near-miss from the state of the object and determine the content of the request based on the estimation result.   The travel recording apparatus according to claim 6, wherein the request unit is configured to estimate a nearby object closest to the host vehicle as a cause object of a near-miss.   The travel recording apparatus according to any one of claims 1 to 7, wherein the voice recognized by the voice recognition unit as the utterance content of the occupant is limited to words and negative words related to near-miss. A primary determination unit (16) for determining whether or not the travel information recorded in the travel information recording unit is a near-miss scene from a state change of the vehicle during vehicle travel;
The request unit is configured to request an explanation of a situation from an occupant when the primary determination unit determines that the traveling information is a near-miss scene. The travel recording device according to item. The near-miss determination unit obtains a score representing near-miss likeness based on the utterance content of the occupant recognized by the voice recognition unit, and the score and the primary determination unit determine the near-miss scene from the state change of the vehicle. The travel recording device according to claim 9, wherein the travel recording device is configured to determine whether or not the travel information recorded in the travel information recording unit is a near-miss scene based on a weighted sum obtained by weighting and adding the score used for the calculation. . Information on near-miss scenes obtained by acquiring near-miss information from a recording medium in which text information representing travel information including images around the vehicle and utterances of passengers is recorded as near-miss information that occurred during vehicle travel A browsing device (50),
An extraction unit (52) for extracting near-miss information corresponding to an externally designated query from the recording medium (40);
A display unit (54) for displaying a near-miss scene based on the near-miss information extracted by the extraction unit;
With
The browsing device configured to display a list of utterance contents and the frequency of occurrence of utterance contents based on the text data of the utterance contents included in the near-miss information extracted by the extraction section. Information on near-miss scenes obtained by acquiring near-miss information from a recording medium in which text information representing travel information including images around the vehicle and utterances of passengers is recorded as near-miss information that occurred during vehicle travel A browsing device,
An extraction unit for extracting near-miss information corresponding to a query designated from the outside from the recording medium;
Based on near-miss information extracted by the extraction unit, a display unit that displays a near-miss scene,
With
The browsing device configured to be capable of extracting information including text data of utterance contents corresponding to the query by designating the query. Information on near-miss scenes obtained by acquiring near-miss information from a recording medium in which text information representing travel information including images around the vehicle and utterances of passengers is recorded as near-miss information that occurred during vehicle travel A browsing device,
An extraction unit for extracting near-miss information corresponding to a query designated from the outside from the recording medium;
Based on near-miss information extracted by the extraction unit, a display unit that displays a near-miss scene,
With
The display device is configured to display the content of the utterance based on text data included in the near-miss scene from the near-miss information extracted by the extraction unit.