Train security system, voice anomaly detection device, and voice anomaly detection program

Abstract

To automatically detect an emergency inside cars and notify the crew.SOLUTION: A train security system (1) is a security system mounted on a train (T) with multiple cars, and includes: voice input means (11) installed inside each car (C1, C2, ...); detection means (12) for detecting voice abnormalities in cars by recognizing special voice including at least one of screaming, rescue voice, and noise based on voice data inputted in the voice input means; and notifying means for causing an emergency in the cars to be notified of based on voice abnormality detection result by the detection means.SELECTED DRAWING: Figure 1

Description

Technical Field

[0006] , , , , , , , ,

[0005]

[0001] The present invention relates to a train security system, a voice abnormality detection device, and a voice abnormality detection program mounted on a train having a plurality of vehicles.

Background Art

[0002] For example, as disclosed in Japanese Patent Application Laid-Open No. 2009-124904 (Patent Document 1), an abnormality notification system that detects an abnormality in a railway vehicle itself and notifies the crew of the railway vehicle of the abnormality has been conventionally proposed. Also, as disclosed in Japanese Patent No. 3888932 (Patent Document 2), an image transmission device for vehicle monitoring intended to notify a passenger of the occurrence of an emergency situation by operation has been conventionally proposed.

[0003] Furthermore, Japanese Patent Application Laid-Open No. 2017-145066 (Patent Document 3) discloses a lift device provided with a movable part emergency stop means for emergency stopping the movable part when a scream is recognized based on a signal from a scream recognition module.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Patent Document 2

Patent Document 3

Summary of the Invention

Problems to be Solved by the Invention

[0005] Recently, serious incidents have occurred frequently inside the vehicles of railway trains. Although emergency buttons are installed inside the vehicles, it is difficult to notify the crew of an emergency situation because there is no time to press the emergency button under urgent circumstances.

[0006] Furthermore, while there are security systems in place that install security cameras inside train cars and allow crew members to view the footage on display devices installed in the conductor's compartment, the more cameras there are, especially in trains with many cars, the more difficult it becomes for crew members to review all the footage, increasing the risk of overlooking emergencies.

[0007] Therefore, simply displaying in-vehicle footage on a display device in the conductor's compartment or elsewhere may not allow crew members to immediately grasp the occurrence of an emergency.

[0008] The present invention was made to solve the above-mentioned problems, and its objective is to provide a train security system, an audio anomaly detection device, and an audio anomaly detection program that can automatically detect an emergency situation inside a train and notify the crew. [Means for solving the problem]

[0009] A train security system according to one aspect of this invention is a security system installed on a train having multiple carriages, comprising: an audio input means installed inside each carriage; a detection means for detecting an audio anomaly inside the carriage by recognizing special sounds, including at least one of screams, rescue sounds, and noise, based on audio data input to the audio input means; and a notification means for notifying an emergency situation inside the carriage based on the audio anomaly detection result by the detection means.

[0010] Preferably, a voice anomaly detection device is installed in each vehicle, which integrally includes a main unit containing detection means and voice input means.

[0011] Preferably, multiple audio anomaly detection devices are installed in each train car. In this case, it is desirable that the train security system further includes a train car anomaly determination means that determines whether or not there is an audio anomaly on a train car basis based on the audio anomaly detection results of each audio anomaly detection device.

[0012] Preferably, the train security system further includes a relay device that is connected to each audio anomaly detection device via a network, receives audio anomaly detection results on a device-by-device basis from each audio anomaly detection device, and sends back the received multiple audio anomaly detection results to the audio anomaly detection device. In this case, it is desirable that the vehicle anomaly determination means is implemented by the audio anomaly detection device.

[0013] Preferably, the train security system further includes a running noise reduction means provided between the voice input means and the detection means. The running noise reduction means preferably includes an estimation processing means for estimating running noise from the input voice input to the voice input means, and a reduction processing means for removing running noise from the input voice based on the estimation result by the estimation processing means.

[0014] Preferably, the train security system further includes acquisition means for acquiring train information, including train speed information. In this case, the running noise reduction means may include means for changing parameters used by the estimation processing means for estimating running noise according to the train information.

[0015] Preferably, each train car is equipped with a video camera, and the notification means includes a display means for displaying the video footage from the video camera installed in the car in which the audio abnormality was detected.

[0016] Furthermore, since most trains have emergency buttons installed in each car, the notification method may also include a means of notifying of an emergency by linking the emergency button installed in the car in which the audio abnormality is detected.

[0017] Another aspect of this invention relates to a voice anomaly detection device mounted on a train car, comprising: a voice input means installed inside the car; a detection means for detecting a voice anomaly inside the car by recognizing special sounds, including at least one of screams, rescue sounds, and noise, based on voice data input to the voice input means; and a communication means for notifying the outside of result information, including the voice anomaly detection result by the detection means.

[0018] The voice abnormality detection program according to another aspect of the present invention causes a computer to execute steps of inputting voice data from voice input means installed inside a vehicle of a train, and detecting a voice abnormality inside the vehicle by recognizing special voice including at least one of a scream, a rescue voice, and a commotion based on the input voice data, and a step of notifying an external device of result information including a voice abnormality detection result by the detection step.

Advantages of the Invention

[0019] According to the present invention, an emergency situation inside a vehicle can be automatically detected and notified to a crew member.

Brief Description of the Drawings

[0020] [Figure 1] It is a diagram schematically showing a schematic configuration of a train security system according to an embodiment of the present invention. [Figure 2] It is a block diagram showing an overall configuration of a train security system according to an embodiment of the present invention and a schematic configuration of a train information management system. [Figure 3] (A) and (B) are block diagrams showing a configuration example of a voice abnormality detection device according to an embodiment of the present invention. [Figure 4] (A) and (B) are block diagrams showing a configuration example of a server according to an embodiment of the present invention. [Figure 5] (A) and (B) are block diagrams showing a configuration example of a control device of a train information management system according to an embodiment of the present invention. [Figure 6] (A) to (D) are diagrams showing a structural example of each transmission data transmitted and received by a server in an embodiment of the present invention. [Figure 7] It is a flowchart showing an overall operation of a train security system according to an embodiment of the present invention. [Figure 8](A) and (B) are flowcharts showing the speech recognition process and the device-level anomaly determination process, respectively, performed by the speech anomaly detection device according to an embodiment of the present invention. [Figure 9] (A) is a flowchart showing the vehicle-level abnormality determination process performed by the voice abnormality detection device according to an embodiment of the present invention, and (B) is a diagram showing an example of priority calculation performed in the vehicle-level abnormality determination process. [Figure 10] This flowchart shows the notification determination process executed by the control device of the train information management system according to an embodiment of the present invention. [Figure 11] (A) and (B) are schematic diagrams showing examples of screens displayed in S313 of Figure 10. [Modes for carrying out the invention]

[0021] Embodiments of the present invention will be described in detail with reference to the drawings. In the drawings, the same or corresponding parts are denoted by the same reference numerals, and their descriptions will not be repeated.

[0022] <Outline configuration> Referring to Figure 1, the schematic configuration of the train security system 1 according to this embodiment will be described. Figure 1 is a schematic diagram showing the schematic configuration of the train security system 1 according to this embodiment.

[0023] The train security system 1 is installed on a train T that has multiple cars, such as a Shinkansen. In this embodiment, for ease of understanding, it is assumed that train T has eight cars, C1, C2, ..., C8, from car 1 to car 8. When the direction of travel is uphill (indicated by arrow A1), car 1 C1 is the leading car, and when the direction of travel is downhill (indicated by arrow A2), car 8 C8 is the leading car.

[0024] In the following explanation, unless there is a need to distinguish between vehicles C1, C2, ..., C8, they will be collectively referred to as vehicle C. The interior space of each vehicle C consists of a passenger compartment R and decks D1 and D2 located at the front and rear of it.

[0025] The train security system 1 comprises at least one voice anomaly detection device 10 installed in each train car C, and a server 20. The server 20 is connected via a network to all voice anomaly detection devices 10 and the train information management system SYS. The train information management system SYS is a system that has the functions necessary for the operation of train T. The train information management system SYS manages train information including speed information and mileage information, and typically has the function of acquiring and displaying video footage from video cameras installed in train cars C. In this embodiment, the train security system 1 works in cooperation with the train information management system SYS to detect emergencies in train cars C and notify the crew.

[0026] The audio anomaly detection device 10 includes at least one microphone unit 11 and a main unit 12. The main unit 12 detects audio anomalies in the vehicle by recognizing "special sounds," which include at least one of screams, rescue sounds, and noise, based on audio data from the microphone unit 11. Special sounds refer to special sounds that are different from normal conversation sounds and are emitted in emergency situations.

[0027] In this embodiment, the microphone unit 11 and the main unit 12 are integrated. Therefore, compared to a configuration where the microphone unit 11 and the main unit 12 are separate, it is possible to prevent a decrease in recognition performance due to audio signal transmission errors and an increase in transmission capacity.

[0028] As shown in Figure 1, considering the range of sound that can be picked up by a single microphone unit 11 and the performance differences due to the direction of travel (uphill, downhill), it is desirable to have sound anomaly detection devices 10 on both the front and rear sides of vehicle C. In other words, it is desirable to have multiple sound anomaly detection devices 10 (typically two) in each vehicle C. In this embodiment, sound anomaly detection devices 10 are provided at the boundary between the passenger compartment R and the front deck D1, and at the boundary between the passenger compartment R and the rear deck D2. This improves the accuracy of sound anomaly detection in each sound anomaly detection device 10. The two sound anomaly detection devices 10 provided in each vehicle C are connected to the server 20, for example, via an Ethernet®-compatible optical transmission device and hub. Each sound anomaly detection device 10 detects the presence or absence of a sound anomaly and notifies the server 20 of the detection result. In other words, it notifies the server 20 of the result information, including the sound anomaly detection result.

[0029] As shown in Figure 1, if there is not a one-to-one relationship between vehicle C and the audio anomaly detection device 10, for example, if we assume that an audio anomaly occurs in vehicle C2, then not only the audio anomaly detection devices 10 located in front of and / or behind vehicle C2, but also the audio anomaly detection devices 10 located behind the preceding vehicle C1 or in front of the following vehicle C3 (hereinafter referred to as "adjacent positions of the preceding and succeeding vehicles") may detect the audio anomaly. In this case, if we determine which vehicle is abnormal based solely on the presence or absence of an audio anomaly, there is a possibility that vehicles C1 or C3, which are not actually experiencing an emergency, may also be mistakenly identified as abnormal vehicles.

[0030] Therefore, the train security system 1 performs a two-stage determination: first, it detects audio anomalies for each audio anomaly detection device 10, and second, it determines whether there is an abnormal vehicle (a vehicle where an emergency may have occurred) based on the detection results of the audio anomaly for each device and the installation location information. In other words, the first determination detects audio anomalies at the device level, and the second determination detects audio anomalies at the vehicle level.

[0031] In this embodiment, the server 20 functions as a relay device, and both primary and secondary determinations are performed in the audio anomaly detection device 10. The server 20 aggregates the audio anomaly detection results (primary determination results) from all audio anomaly detection devices 10 on a device-by-device basis and returns them to each audio anomaly detection device 10, thereby enabling each audio anomaly detection device 10 to detect audio anomalies on a vehicle-by-vehicle basis (performing a secondary determination).

[0032] Server 20 notifies the train information management system SYS of the detection results for audio anomalies (presence or absence of an abnormal vehicle) on a vehicle-by-vehicle basis. As a result, the notification means (for example, the display device 120 described later) equipped in the train information management system SYS is notified of the information (for example, the vehicle number) of vehicle C which has been determined to be an abnormal vehicle, so that the emergency situation inside vehicle C can be notified to the crew, or to the ground operations manager in vehicles equipped with ground-to-vehicle communication functions.

[0033] <System Configuration> Referring to Figure 2, a specific configuration example of the train security system 1 of this embodiment will be described. Figure 2 is a block diagram showing the overall configuration of the train security system 1 of this embodiment and the schematic configuration of the train information management system SYS.

[0034] The train information management system SYS comprises a control unit (hereinafter referred to as the "management control unit") 100 that controls the entire system, a vehicle control unit (CTL) 110 and a display device 120 provided for each vehicle C. It also includes at least one video camera 130 in each vehicle C. The management control unit 100 is connected to the vehicle control units 110, the display device 120, and the video cameras 130 via a network. Each vehicle control unit 110 is connected to an emergency button installed in each vehicle C. When an emergency button is pressed, an emergency notification is sent from the vehicle control unit 110 to the management control unit 100. In this case, the management control unit 100 notifies the crew of the emergency by performing notification processing, such as generating an emergency sound.

[0035] The display device 120 is installed, for example, in the driver's cab of the leading car, and the video from the video camera 130 is displayed by the management control device 100. The display device 120 may also be installed in the driver's cab of the trailing car, etc. The display device 120 may be implemented, for example, by a general-purpose computer with a display, or the management control device 100 and the display device 120 may be provided as an integrated unit.

[0036] As shown in Figure 2, the server 20 transmits and receives data (SDRa, SDa) with the management control device 100 at regular intervals, and also transmits and receives data (SDRb, SDb) with each voice anomaly detection device 10 at regular intervals. In response to the request data SDRa from the management control device 100, the server 20 returns reply data SDa to the management control device 100, which includes the contents of the data SDb received from each voice anomaly detection device 10. The server 20 simultaneously transmits request data SDRb, which reflects the data SDRa from the management control device 100, to all voice anomaly detection devices 10, and receives reply data SDa from all voice anomaly detection devices 10. An example of the structure of data SDRa, SDa, SDRb, and SDb shown in Figure 2 will be described in detail later.

[0037] The data transmission period between the server 20 and the management control device 100 is typically set to be shorter than the data transmission period between the server 20 and the voice anomaly detection device 10. For example, the former is 0.2 seconds and the latter is 1.0 second. However, both periods may be the same.

[0038] <Example of a voice anomaly detection device configuration> Referring to Figure 3, an example of the configuration of the voice anomaly detection device 10 according to this embodiment will be described.

[0039] Figure 3(A) is a block diagram showing the hardware configuration of the voice anomaly detection device 10. The voice anomaly detection device 10 consists of at least one microphone unit 11 and a main unit 12. The microphone unit 11 functions as a voice input means. The main unit 12 functions as a detection means and includes a processor 13 that performs various calculations, a memory 14 that stores various data and programs, and a communication I / F (interface) 15 for communicating with the server 20.

[0040] Since the door between the passenger compartment R and the decks D1 and D2 of vehicle C is normally closed, it is desirable that the microphone unit 11 be provided in both the passenger compartment R and the decks D1 and D2. Specifically, the microphone unit 11 is provided on both sides of the partition wall with the door. Alternatively, it may be provided in the ceiling on the passenger compartment R side and the ceiling on the deck D1 or D2 side near this partition wall. This allows the voice anomaly detection device 10 installed at the front of each vehicle C to accurately detect sound in the relatively forward area of ​​the passenger compartment R and the entire area of ​​the leading deck D1. Similarly, the voice anomaly detection device 10 installed at the rear of each vehicle C can accurately detect sound in the relatively rear area of ​​the passenger compartment R and the entire area of ​​the rear deck D2.

[0041] Memory 14 pre-stores the IP addresses and installation location information of all audio anomaly detection devices 10, including its own device. Specifically, the configuration file in memory 14 stores the IP address and installation location information (e.g., in front of car 1) associated with the identification ID of each audio anomaly detection device 10. It is desirable that the information in the configuration file in memory 14 be updatable by an external device (such as a maintenance PC) not shown in the diagram. This allows for replacement of audio anomaly detection devices 10 due to malfunction or other reasons.

[0042] Figure 3(B) is a block diagram showing the functional configuration of the main unit 12 of the voice anomaly detection device 10. The main unit 12 includes a voice recognition unit 31 for recognizing special voices from voices input from two microphone units 11, a request acquisition unit 32 for acquiring request data SDRb from the server 20, and a reply processing unit 33 for generating reply data SDb in response to the request data SDRb.

[0043] The voice recognition unit 31 preferably has a scream recognition function to recognize screams and a rescue voice recognition function to recognize rescue voices, as well as a noise detection function to detect "noise" that is a mixture of screams and rescue voices. In other words, the voice recognition unit 31 detects screams, rescue voices, and noise as special sounds from the input voice.

[0044] "Scream" is defined as sound that exceeds a predetermined volume level, is sufficiently loud compared to ambient noise, and contains frequency components higher than speech sounds, and persists for a certain period of time. "Rescue voice" is sound that is detected by word recognition or verbatim recognition as a phrase requesting rescue (e.g., help, danger). "Noise" is defined as sound that persists for a certain period of time and frequently exceeds the volume level of normal conversation.

[0045] In this case, while train T is in motion, the audio input from the microphone unit 11 includes running noise. Therefore, if the input audio is analyzed directly, there is a risk that the special sounds described above may not be detected accurately. For this reason, the audio recognition unit 31 in this embodiment has a running noise removal function preceding each recognition function. This makes it possible to minimize the influence of running noise, which changes moment by moment.

[0046] The request acquisition unit 32 acquires request data SDRb from the server 20 via the communication interface 15. This request data SDRb, like the request data SDRa from the management control device 100, includes train information and clear information described later. The request data SDRb also includes the previous audio anomaly detection results for each device from all audio anomaly detection devices 10.

[0047] The reply processing unit 33 includes a voice anomaly detection unit 34 that detects the presence or absence of special sounds (presence or absence of voice anomalies) based on the most recent recognition result by the voice recognition unit 31. Specifically, the voice anomaly detection unit 34 determines the presence or absence of screams, rescue sounds, and noise. The reply processing unit 33 also includes a vehicle anomaly determination unit 35 that determines the presence or absence of voice anomalies on a vehicle C basis based on the previous device-level voice anomaly detection result included in the request data SDRb.

[0048] The reply processing unit 33 generates reply data SDb, which includes the judgment results from the voice anomaly detection unit 34 and the vehicle anomaly determination unit 35. The generated reply data SDb is then sent back to the server 20 via the communication interface 15.

[0049] Each functional unit shown in Figure 3(B) is typically implemented by the processor 13 executing software. The speech recognition unit 31 is preferably configured as a dedicated module.

[0050] <Example Server Configuration> Referring to Figure 4, an example of the configuration of the server 20 according to this embodiment will be described.

[0051] Figure 4(A) is a block diagram showing the hardware configuration of the server 20. The server 20 includes a processor 21 that performs various calculations, a memory 22 that stores various data and programs, a non-volatile storage device 23 such as a hard disk, a communication interface 24 for communicating with the management control device 100, and a communication interface 25 for communicating with each audio anomaly detection device 10.

[0052] Figure 4(B) is a block diagram showing the functional configuration of the server 20. The server 20 includes a request acquisition unit 41 that acquires request data SDRa from the management control device 100, a reply processing unit 42 that generates reply data SDa in response to the request data SDRa and sends it back to the management control device 100, a request transmission unit 43 that generates request data SDRb and sends it to each voice anomaly detection device 10, and a reply receiving unit 44 that receives reply data SDb from each voice anomaly detection device 10 and outputs it to the reply processing unit 42.

[0053] The request transmission unit 43 includes an update unit 45 that updates train information etc. using the request data SDRa acquired by the request acquisition unit 41, and transmits the updated data from the update unit 45 as request data SDRb to each voice anomaly detection device 10.

[0054] The reply processing unit 42 includes an update unit 46 that updates the vehicle abnormality determination result based on the reply data SDb received by the reply receiving unit 44, and sends the updated data from the update unit 46 back to the management control device 100 as reply data SDa.

[0055] Each of the functional units shown in Figure 4(B) is typically realized by the processor 21 executing software.

[0056] <Example of a control system configuration> Referring to Figure 5, an example of the configuration of the management control device 100 according to this embodiment will be described.

[0057] Figure 5(A) is a block diagram showing the hardware configuration of the management control device 100. The management control device 100 includes a processor 101 that performs various calculations for operating train T, a memory 102 that stores various data and programs, a non-volatile storage device 103 such as a hard disk, a communication interface 104 for communicating with a vehicle control device 110 provided for each vehicle C, an input / output interface 105 connected to a display device 120, and an input / output interface 106 connected to a video camera 130. In this embodiment, it further includes a communication interface 107 for communicating with a server 20.

[0058] Figure 5(B) is a block diagram showing the functional configuration of the management control device 100 that works in conjunction with the train security system 1. The management control device 100 includes a train information acquisition unit 51 that (constantly) acquires train information based on information from the vehicle control device 110, a request transmission unit 52 that generates request data SDRa including the acquired train information and sends it to the server 20, a reply receiving unit 53 that receives reply data SDa from the server 20, and a notification processing unit 55 that performs notification processing based on the reply data SDa.

[0059] The reply receiving unit 53 includes a determination unit 54 that determines whether or not a vehicle has experienced an audio anomaly, based on the vehicle anomaly determination result contained in the reply data SDa, and outputs the determination result to the notification processing unit 55. The notification processing unit 55, for example, notifies (displays) the vehicle number of the vehicle C that was determined to be abnormal on the display device 120, and also displays a video of the vehicle C that was determined to be abnormal. This allows the crew to easily check the situation inside vehicle C.

[0060] Each of the functional units shown in Figure 5(B) is typically realized by the processor 101 executing software.

[0061] <Examples of the structure of each transmitted data> Referring to Figure 6, an example of the structure of each transmission data transmitted and received by the server 20 in this embodiment will be described.

[0062] Figure 6(A) shows an example of the data structure of the request data SDRa sent from the management control device 100 to the server 20. The request data SDRa includes a header I11, train information I12, and clear information I13.

[0063] Train information I12 includes at least speed information and mileage information, and further includes broadcast information and door opening / closing information.

[0064] The clear information I13 is set to "1" only when an abnormal condition is cleared. The request transmission unit 52 of the management control device 100 sets the clear information to "1" when the user (crew member) gives an instruction to clear the abnormal condition after the notification processing unit 55 has performed the notification processing.

[0065] Figure 6(B) shows an example of the data structure of the request data SDRb sent from the server 20 to each audio anomaly detection device 10. The request data SDRb includes a header I21, train information I22, clear information I23, and the audio anomaly detection result I24 for all devices.

[0066] Train information I22 and clear information I23 are the same as train information I12 and clear information I13 shown in Figure 6(A).

[0067] The total audio anomaly detection result I24 for all devices includes the determination results from the audio anomaly detection units 34 of all audio anomaly detection devices 10, i.e., the determination results for each device. In this embodiment, it includes the determination result I101 of the audio anomaly detection device 10 located at the front of car 1, the determination result I102 of the audio anomaly detection device 10 located at the rear of car 1, ... and the determination result I116 of the audio anomaly detection device 10 located at the rear of car 8.

[0068] The judgment results I101, I102, ..., I116 of each sound anomaly detection device 10 include a scream flag I121 indicating the presence or absence of screams, a noise flag I122 indicating the presence or absence of commotion, and a rescue flag I123 indicating the presence or absence of rescue sounds. The initial state of flags I121 to I123 is "0", and it transitions to "1" when the corresponding sound is detected.

[0069] Figure 6(C) shows an example of the data structure of the reply data SDb sent from each audio anomaly detection device 10 to the server 20. The reply data SDb includes a header I31, a master unit flag I32, the audio anomaly detection result I33 of the device itself, and the vehicle anomaly determination result I34 of the device itself. The identification ID of the device itself is recorded in this header I31.

[0070] The master unit flag I32 indicates whether the device is the master unit, and is set to "1" only if it is the master unit. In this embodiment, for example, the voice abnormality detection device 10 located at the front in the direction of travel is set as the master unit.

[0071] The device's audio anomaly detection result I33 indicates the determination result by the device's audio anomaly detection unit 34, and specifically includes a scream flag I201 indicating the presence or absence of screams, a noise flag I202 indicating the presence or absence of commotion, and a rescue flag I203 indicating the presence or absence of rescue sounds.

[0072] The vehicle abnormality determination result I34 of the device indicates the determination result by the vehicle abnormality determination unit 35 of the device, and specifically includes a vehicle abnormality flag I211 indicating whether or not there is an abnormality in vehicle C1 of vehicle C1, a vehicle abnormality flag I212 indicating whether or not there is an abnormality in vehicle C2 of vehicle C2, ..., and a vehicle abnormality flag I218 indicating whether or not there is an abnormality in vehicle C8 of vehicle C8.

[0073] Figure 6(D) shows an example of the data structure of the reply data SDa sent from the server 20 to the management control device 100. The reply data SDa includes a header I41 and a vehicle abnormality determination result I42 sent from the master unit.

[0074] Vehicle abnormality detection result I42 is the same as vehicle abnormality detection result I34, which is included in the reply data SDb from the voice abnormality detection device 10 to the server 20, where the master unit flag I22 indicates "1". Note that this vehicle abnormality detection result I34 is basically the same for voice abnormality detection devices 10 other than the master unit.

[0075] <Train security system operation> Referring to Figure 7, the overall operation of the train security system 1 according to this embodiment will be described.

[0076] Figure 7 is a flowchart illustrating the operation of the train security system 1. In Figure 7, the step numbers for processes executed by the voice anomaly detection device 10 are shown in the 100s, the step numbers for processes executed by the server 20 are shown in the 200s, and the steps for processes executed by the train information management system SYS (management control device 100) are shown in the 300s. In addition, Figure 7 shows an example where the data transmission period between the management control device 100 and the server 20 is the same as the data transmission period between the voice anomaly detection device 10 and the server 20, in order to easily understand the overall flow of the train security system 1. In an actual system, the train information management system SYS, the voice anomaly detection device 10, and the server 20 may operate asynchronously.

[0077] Each audio anomaly detection device 10 performs a startup process when power is supplied in conjunction with the pressurization of train T (step S101 (hereinafter, "step S" will be abbreviated as "S")). The server 20 also performs a startup process in the same manner (S201). The train information acquisition unit 51 of the management control device 100 starts acquiring train information.

[0078] Each voice anomaly detection device 10 starts voice recognition processing after the startup process is completed (S103). The voice recognition processing is explained by listing the subroutines in Figure 8(A).

[0079] Referring to Figure 8(A), when the voice recognition unit 31 of the voice anomaly detection device 10 receives voice input from the microphone unit 11 (S121), it first performs a process to remove running noise. Specifically, it constantly estimates the running noise (S122) and then performs a process to remove the estimated running noise from the input voice (S123). The reason for constantly estimating the running noise in this way is that the running noise changes sequentially depending on the installation location, the running speed of the train T, the running environment of the train T (inside or outside of tunnels, etc.), weather, etc., making prior learning difficult.

[0080] In S122, it is desirable to first determine whether or not the input audio contains human voice features, and then perform noise estimation on a frame-by-frame basis for input audio that does not contain human voice features. Then, in S123, if the input audio does not contain human voice features, the noise estimated in S122 (driving noise) is removed from the input audio. On the other hand, if it is determined that the input audio contains human voice features, the noise (i.e., the noise estimated immediately before that does not contain human voices) is removed from the input audio using the estimation result from the previous frame. By doing this, it is possible to prevent special sounds that are the target of recognition (such as screams and rescue calls) from being removed as noise.

[0081] The voice recognition unit 31 performs scream recognition processing (S124), noise detection processing (S125), and rescue recognition processing (S126) in parallel on the audio from which driving noise has been removed. These recognition processes (detection processes) are executed continuously, and the results of each recognition are stored chronologically in memory such as RAM.

[0082] Referring again to Figure 7, the request transmission unit 52 of the management control device 100 generates the initial request data SDRa-1 based on the current train information and sends it to the server 20 (S301). The clear information I13 in the initial request data SDRa-1 is the initial value ("0").

[0083] When the request acquisition unit 41 of the server 20 receives request data SDRa-1 from the management control device 100 (S203), it reads train information I12 and clear information I13 from the request data SDRa-1 and temporarily records them in internal memory.

[0084] The request transmission unit 43 of the server 20 generates the initial request data SDRb-1, which updates the train information I22 and clear information I23 according to the train information I12 and clear information I13 temporarily recorded in the internal memory, and transmits it simultaneously to all audio anomaly detection devices 10 (S205). The audio anomaly detection result I24 in the initial request data SDRb-1 is the initial value (all "0").

[0085] When the request acquisition unit 32 of each audio anomaly detection device 10 receives the initial request data SDRb-1 (S105), it determines whether its device is the master unit or not based on the train information (specifically, the direction of travel of train T) contained in the request data SDRb-1 (S107).

[0086] Next, the voice anomaly detection unit 34 of the reply processing unit 33 performs an anomaly determination process for the device unit (itself) (S108). This process is explained by showing the subroutines in Figure 8(B).

[0087] Referring to Figure 8(B), the voice anomaly detection unit 34 acquires the recognition results from the voice recognition unit 31, that is, the recognition results stored in time series in S103 (S131). Then, if the recognition type is screaming or noise, it calculates a moving average for each type (S133) and determines whether or not there is a voice anomaly (S135). Specifically, it determines whether or not there is screaming or noise. The presence or absence of rescue sounds is determined by the recognition results of the voice recognition unit 31.

[0088] Referring again to Figure 7, the reply processing unit 33 generates reply data SDb-1 that reflects the judgment result of S135 and sends it back to the server 20 (S109). Specifically, it generates reply data SDb-1 by changing the values ​​of the recognition types that are determined to be "present" among the scream flag I121, noise flag I122, and rescue flag I123 of the sound anomaly detection result I33 from the initial value "0" to "1". Also, if it is determined to be a master unit in S107, the master unit flag I32 of the reply data SDb-1 is set to "1". The vehicle anomaly judgment result I34 in the initial reply data SDb-1 remains at its initial value (all "0").

[0089] When the reply receiving unit 44 of the server 20 receives reply data SDb-1 from all voice anomaly detection devices 10 (S207), it temporarily records the voice anomaly detection results I33 from all voice anomaly detection devices 10 and the vehicle anomaly detection result I34 from the voice anomaly detection device 10 that was determined to be the master unit in its internal memory. The voice anomaly detection results I33 included in the reply data SDb-1 differ from device to device.

[0090] The reply processing unit 42 of the server 20 updates the vehicle abnormality determination result I42 of the reply data SDa-1 with the vehicle abnormality determination result I34 (i.e., the latest determination result) temporarily stored in the internal memory, and sends it back to the management control device 100 as reply data for the request data SDRa-1 received in S203 (S209).

[0091] When the reply receiving unit 53 of the management control device 100 receives the reply data SDa-1 (S303), the determination unit 54 performs a notification determination (S304). Since the vehicle abnormality determination result I42 of the initial reply data SDa-1 is an initial value (all "0"), the notification determination process will be described later.

[0092] The control device 100 then generates a second request data SDRa-2 based on the current train information and sends it to the server 20 (S305). The clear information I13 in the second request data SDRa-2 also remains at its initial value.

[0093] When the request acquisition unit 41 of the server 20 receives request data SDRa-2 from the management control device 100 (S211), it reads train information I12 and clear information I13 from the request data SDRa-2 and temporarily records them in internal memory.

[0094] The request transmission unit 43 of the server 20 updates the train information I22 and clear information I23 in accordance with the train information I12 and clear information I13 obtained from the management control device 100 in S211, and generates a second request data SDRb-2 in which the voice anomaly detection result I24 is updated in accordance with the voice anomaly detection result I33 obtained from all voice anomaly detection devices 10 in S207 (S213). That is, the flags I121 to I123 of all judgment results I101 to I116 included in the voice anomaly detection result I33 shown in Figure 6(B) are updated to the values ​​of the flags I201 to I203 of the voice anomaly detection result I33 of the previous (first in this example) reply data SDb-1. The request transmission unit 43 then simultaneously transmits the request data SDRb-2, which includes the judgment results I101 to I116 (i.e., voice anomaly detection result I24) from all voice anomaly detection devices 10, to all voice anomaly detection devices 10.

[0095] When the request acquisition unit 32 of each audio anomaly detection device 10 receives the second request data SDRb-2 (S111), the audio anomaly detection unit 34 performs an anomaly determination process for the device unit (S112) and also performs an anomaly determination process for the vehicle unit (S113). The anomaly determination process for the device unit is the same as in step S108. The anomaly determination process for the vehicle unit is explained by showing the subroutines in Figure 9(A).

[0096] Referring to Figure 9(A), the vehicle abnormality determination unit 35 first reads the determination results I101 to I116 (i.e., the voice abnormality detection result I24) from the request data SDRb-2 by all voice abnormality detection devices 10 (S141). Referring to the determination results I101 to I116, the vehicle abnormality determination unit 35 identifies the installation location of the voice abnormality detection device 10 that detected the voice abnormality, i.e., the car number and location (passenger compartment, front deck, rear deck) of vehicle C (S143). The presence or absence of a voice abnormality detection device 10 that detected a voice abnormality can be determined by the values ​​of flags I201 to I203 ("1" or "0"). If any of the following are detected: screams, noise, or rescue sounds, it is determined that there is a voice abnormality.

[0097] Next, if an audio abnormality is detected at an adjacent position between the front and rear vehicles, the vehicle abnormality determination unit 35 performs a priority determination based on at least one of the audio type and sound pressure (S145). This priority determination may be performed only if the audio abnormality detection device 10 located at an adjacent position among two vehicles C adjacent in the front-rear direction detects an audio abnormality.

[0098] Figure 9(B) shows an example of priority calculation in S145. For example, if the scream flag is "1", the points are set to "4", the noise flag is "1", the points are set to "2", the rescue flag is "1", and the points are set to "1" when the sound pressure is above a predetermined value, the points for two adjacent sound anomaly detection devices 10 are calculated accordingly. As a result of the calculation, vehicle C, which has the sound anomaly detection device 10 with the higher point count, is determined to be the abnormal vehicle (S147). This reduces the burden on cabin crew in subsequent notification processing (the burden of confirming sound anomalies that appear to be caused by the same factor).

[0099] Furthermore, in S147, if the audio anomaly detection device 10, which is not related to the specific conditions of the adjacent location, detects any of the following: screams, noise, and rescue sounds, the vehicle C on which the device is installed may be determined to be an abnormal vehicle.

[0100] Referring again to Figure 7, once the device-level anomaly detection process (S112) and the vehicle-level anomaly detection process (S113) are completed, the reply processing unit 33 generates reply data SDb-2 that reflects these detection results and sends it back to the server 20 (S115). Of the anomaly flags I211 to I218 for vehicles 1 to 8 in the vehicle anomaly detection result I34 included in the reply data SDb-2, the flag for the vehicle number determined to be an anomaly in S147 is set to "1" and sent to the server 20.

[0101] When the server 20's reply receiving unit 44 receives reply data SDb-2 from all the voice anomaly detection devices 10 (S215), it overwrites the contents of the previous (1st) reply data SDb-1 recorded in the internal memory with the contents of the current (2nd) reply data SDb-2. In this way, the latest voice anomaly detection results I33 from all the voice anomaly detection devices 10 and the latest vehicle anomaly detection results I34 from the voice anomaly detection device 10 determined to be the master unit are temporarily recorded in the internal memory.

[0102] The reply processing unit 42 of the server 20 updates the vehicle abnormality determination result I42 of the reply data SDa-2 with the vehicle abnormality determination result I34 (i.e., the latest determination result) temporarily stored in the internal memory, and sends it back to the management control device 100 as reply data for the request data SDRa-2 received in S211 (S217).

[0103] When the reply receiving unit 53 of the management control device 100 receives the reply data SDa-2 (S307), the determination unit 54 performs a notification determination (S308). The notification determination process will now be explained.

[0104] Figure 10 is a flowchart showing the notification determination process executed by the management control device 100. Referring to Figure 10, the determination unit 54 of the management control device 100 reads the vehicle abnormality determination result I42 from the reply data SDa-2 and determines whether or not there is an abnormal vehicle (S311). Specifically, it determines whether any of the abnormality flags for vehicle 1 (I211), vehicle 2 (I212), ..., and vehicle 8 (I218) indicate "1".

[0105] If it is determined that there is no abnormal vehicle (NO in S311), the notification determination process ends and the system returns to the main routine shown in Figure 7. On the other hand, if it is determined that there is an abnormal vehicle (YES in S311), the notification processing unit 55 executes a notification process to notify the emergency situation in vehicle C (S313). Specifically, the notification processing unit 55 notifies the display device 120 (Figure 2) that an audio abnormality has been detected in vehicle C, the vehicle for which the flag is set to "1", for example, by displaying a pop-up screen. An example of the screen display at this time is shown in Figure 11(A).

[0106] In Figure 11(A), a pop-up screen 123 is displayed on the display 121 of the display device 120, which reads, "Audio anomaly detected in car number XX." This display 121 is, for example, the main display of the display device 120. This ensures that the presence of car C, where the audio anomaly has been detected, is reliably notified to the crew in the conductor's compartment or other areas (crew compartment). It is also desirable that a selection button 124 labeled "Check camera footage" is displayed on this screen 123. When this selection button 124 is selected, the notification processing unit 55 displays the video footage from the video camera 130 installed in car C, where the audio anomaly was detected, on the display 121. This allows the crew to immediately check the camera footage of car C.

[0107] Figure 11(B) shows an example of the screen display when the selection button 124 is selected. In this example, the images from all (or more) cameras installed in the target vehicle C are displayed separately on the two displays 121 and 122 of the display device 120. If video cameras 130 are installed in the passenger compartment R, the front deck D1, and the rear deck D2, the conductor's room can check the situation in each area of ​​vehicle C by displaying the images from each area. This allows the crew to quickly take necessary measures, such as rushing to the target vehicle C or informing passengers in other vehicles C of the emergency via the in-car announcement system. Although this example shows the simultaneous display of camera images from all areas, it is not limited to this example; for example, the camera images from each area may be sequentially switched and displayed on a single display.

[0108] Referring again to Figure 7, the control device 100 generates a third request data SDRa-3 based on the current train information after or in parallel with the notification determination process in S308 and sends it to the server 20 (S309). As a result, in the same flow as described above, the server 20 receives the request data SDRa-3 (S219), the server 20 generates and sends the request data SDRb-3 (S221), and each audio anomaly detection device 10 receives the request data SDRb-3 (S117), and the same process is repeated thereafter.

[0109] Furthermore, the timing at which the management control device 100 sets the clear information I13 of the request data SDRa to "1" is, for example, when the selection button 124 for "checking camera images" shown in Figure 11(A) is pressed. It is desirable that the audio abnormality detection device 10 maintains the vehicle abnormality judgment result until the clear information I13 becomes "1".

[0110] As described above, according to the train security system 1 of this embodiment, the audio anomaly detection device 10 installed in each train car C detects audio anomalies, making it possible to determine (estimate) which train car C may be experiencing an emergency. Furthermore, by preferentially displaying the video of that train car C on the display device 120 installed in the conductor's compartment or elsewhere, the risk of overlooking an emergency is reduced. Therefore, the safety of the train T can be improved. In addition, since it does not require equipment to simultaneously check the video from all video cameras 130 (existing display devices 120 can handle this), the cost of introducing the train security system 1 can be reduced.

[0111] Furthermore, in this embodiment, the master unit is not fixed, and all audio abnormality detection devices 10 perform vehicle abnormality determination processing. Therefore, even if the master unit fails, it can be replaced with another device. Also, if data from the audio abnormality detection device 10 that has been determined to be the master unit fails to reach the server 20, it is possible to change another device to the master unit.

[0112] Furthermore, since the voice recognition unit 31 of the voice abnormality detection device 10 has a driving noise reduction function, the number of voice abnormality detection devices 10 per vehicle C can be reduced.

[0113] Furthermore, since the vehicle abnormality detection process is performed on the audio abnormality detection device 10, the processing load on the server 20 can be reduced.

[0114] <Variation> The following describes a modified version of the train security system 1 according to this embodiment.

[0115] (1) Since train information is transmitted from the server 20 to the voice anomaly detection device 10 (as request data SDRb), the voice recognition unit 31 may use the current train information to estimate running noise and detect special sounds. Specifically, the parameters used to estimate running noise and the parameters used to extract features of various special sounds may be changed according to at least one of the current running speed and running position. Alternatively, if the train information further includes information about in-car announcements and door opening / closing information, the parameters for running noise and the parameters of each recognition unit may be changed according to this information.

[0116] (2) In this embodiment, when an audio anomaly detection device 10 located adjacent to the front and rear vehicles detects an audio anomaly (special sound), priority determination is made according to the type of audio anomaly and sound pressure. However, the embodiment is not limited to this example. For example, the vehicle C with the larger number of audio anomaly detection devices 10 that have detected an audio anomaly may simply be determined to be the anomaly vehicle.

[0117] (3) In this embodiment, vehicle abnormality determination (vehicle-specific voice abnormality determination) is performed on the voice abnormality detection device 10 side, but vehicle abnormality determination may also be performed on the server 20 side.

[0118] (4) In the case of a vehicle C carrying a group of passengers, the voice anomaly detection device 10 installed in the vehicle C may detect noise, etc., so the voice recognition function of the voice anomaly detection device 10 installed in such a vehicle C may be set to OFF. If the train information includes information on whether or not there are group passengers in each vehicle C, each voice anomaly detection device 10 can stop processing by the voice recognition unit 31 by reading the train information contained in the request data SDRb from the server 20.

[0119] (5) In this embodiment, an example has been described in which only the reply data SDb containing the audio anomaly detection result of the device itself is sent back as reply data to the request data SDRb from the server 20. However, other types of data may also be sent back. For example, if the audio anomaly detection device 10 detects an audio anomaly, it may record the audio data before and after that point in time, and send back the recorded audio data as reply data for another command. In this case, the server 20 may record the received audio data in the non-volatile storage device 23. This makes it possible to analyze the audio data later.

[0120] (6) If the management control device 100 determines that there is an abnormal vehicle, it may be linked to the emergency button installed in the vehicle C and perform the same processing as when the emergency button is pressed. In other words, the notification means for notifying of an emergency situation in vehicle C is not limited to the display device 120 described above, and may also include means for notifying of an emergency situation by linking to the emergency button installed in the vehicle C in which the abnormal sound was detected.

[0121] (7) In this embodiment, the display device 120 as a notification means has been described as a component of the train information management system SYS, but the display device 120 may be provided independently of the train information management system SYS.

[0122] (8) Although it is desirable that the microphone unit 11 and the main unit 12 of the voice abnormality detection device 10 be provided as an integrated unit, the microphone unit 11 may be provided in each vehicle C, and the voice abnormality detection function (detection means) included in the main unit 12 may be performed by a single device.

[0123] (8) The voice anomaly detection method performed by the voice anomaly detection device 10 can also be provided as a program. Such a program can be provided by recording it on an optical medium such as a CD-ROM (Compact Disc-ROM) or a computer-readable non-transitory recording medium such as a memory card. The program can also be provided by downloading it over a network.

[0124] The program according to the present invention may call necessary program modules from among the program modules provided as part of a computer's operating system (OS) in a predetermined sequence and at predetermined timings to execute processing. In that case, the program itself does not contain the above modules and processes in cooperation with the OS. Such a program that does not contain modules may also be included in the program according to the present invention.

[0125] Furthermore, the program according to the present invention may be provided as part of another program. In that case, the program itself does not contain modules included in the other program, and processing is executed in cooperation with the other program. Such a program incorporated into another program may also be included in the program according to the present invention.

[0126] The embodiments disclosed herein should be considered in all respects to be illustrative and not restrictive. The scope of the present invention is indicated by the claims rather than by the foregoing description, and all modifications within the meaning and scope equivalent to the claims are intended to be included. [Explanation of symbols]

[0127] 1 Train security system, 10 Voice anomaly detection device, 11 Microphone unit (voice input means), 12 Main unit (detection means), 20 Server, 31 Voice recognition unit, 32, 41 Request acquisition unit, 33, 42 Reply processing unit, 34 Voice anomaly detection unit, 35 Vehicle anomaly determination unit, 43, 52 Request transmission unit, 44, 53 Reply receiving unit, 45, 46 Update unit, 51 Train information acquisition unit, 54 Determination unit, 55 Notification processing unit, 100 Management control device, 110 Vehicle control device, 120 Display device, 130 Video camera, C, C1~C8 Vehicle, D1 Front deck, D2 Rear deck, R Passenger compartment, SDRa, SDRb Request data, SDa, SDb Reply data, SYS Train information management system, T Train.

Claims

1. A security system installed on a train with multiple carriages, A voice input means installed inside each of the aforementioned vehicles, A detection means for detecting an audio anomaly in the vehicle by recognizing special sounds, including at least one of screams, rescue sounds, and noise, based on audio data input to the aforementioned audio input means, Based on the results of the audio anomaly detection by the aforementioned detection means, a notification means for notifying the emergency situation inside the vehicle, The system includes a driving noise reduction means provided between the voice input means and the detection means, The train security system includes a running noise removal means which comprises an estimation processing means for estimating running noise from audio data input to the audio input means, and a removal processing means for removing running noise from the audio data based on the estimation result by the estimation processing means.

2. The train security system according to claim 1, wherein a voice anomaly detection device, which integrally includes the main unit containing the detection means and the voice input means, is installed in each of the vehicles.

3. Multiple audio anomaly detection devices are installed in each of the vehicles. The train security system according to claim 2, further comprising a vehicle abnormality determination means for determining the presence or absence of a voice abnormality on a vehicle-by-vehicle basis based on the voice abnormality detection result for each of the voice abnormality detection devices.

4. The relay device is further connected to each of the aforementioned voice anomaly detection devices via a network, receives the voice anomaly detection results on a device-by-device basis from each of the aforementioned voice anomaly detection devices, and sends back a plurality of the received voice anomaly detection results to the aforementioned voice anomaly detection devices. The train security system according to claim 3, wherein the vehicle abnormality determination means is realized by the voice abnormality detection device.

5. The system further includes means for acquiring train information, including the speed information of the aforementioned train. The train security system according to claim 1, wherein the running noise reduction means includes means for changing the parameters used for estimating running noise by the estimation processing means according to the train information.

6. Each of the aforementioned train cars is equipped with a video camera. The train security system according to any one of claims 1 to 5, wherein the notification means includes a display means for displaying the video footage from the video camera installed in the vehicle in which the audio abnormality was detected.

7. Each of the aforementioned train cars is equipped with an emergency button. The train security system according to any one of claims 1 to 6, wherein the notification means includes means for notifying an emergency situation by linking with the emergency button installed on the vehicle in which an audio abnormality is detected.

8. A security system installed on a train having multiple cars, A voice input means installed inside each of the aforementioned vehicles, A detection means for detecting an audio anomaly in the vehicle by recognizing special sounds, including at least one of screams, rescue sounds, and noise, based on audio data input to the aforementioned audio input means, The vehicle is equipped with a notification means that notifies of an emergency situation inside the vehicle based on the sound anomaly detection result by the detection means, A train security system in which a voice anomaly detection device, which integrally includes a main unit containing the detection means and the voice input means, is installed in each of the vehicles.

9. A security system installed on a train having multiple cars, A voice input means installed inside each of the aforementioned vehicles, A detection means for detecting an audio anomaly in the vehicle by recognizing special sounds, including at least one of screams, rescue sounds, and noise, based on audio data input to the aforementioned audio input means, The vehicle is equipped with a notification means that notifies of an emergency situation inside the vehicle based on the sound anomaly detection result by the detection means, Each of the aforementioned train cars is equipped with an emergency button. The notification means includes a means for notifying an emergency situation by linking with the emergency button installed on the vehicle in which an audio abnormality is detected, in a train security system.

10. A sound anomaly detection device installed in a train car, A voice input means installed inside the vehicle, A detection means for detecting an audio anomaly in the vehicle by recognizing special sounds, including at least one of screams, rescue sounds, and noise, based on audio data input to the aforementioned audio input means, A communication means for notifying an external party of result information, including the result of detecting an abnormal voice by the aforementioned detection means, The system includes a driving noise reduction means provided between the voice input means and the detection means, The audio anomaly detection device includes an estimation processing means for estimating driving noise from audio data input to the audio input means, and a removal processing means for removing driving noise from the audio data based on the estimation result by the estimation processing means.

11. A computer installed in a train car that detects audio abnormalities inside the car, The steps include inputting audio data from an audio input means installed inside the vehicle, A step of estimating road noise from input audio data, A step of removing the driving noise from the audio data based on the estimation result of the driving noise, A detection step to detect an audio anomaly in the vehicle by recognizing special sounds, including at least one of screams, rescue sounds, and noise, based on audio data from which driving noise has been removed, A voice anomaly detection program that performs the step of notifying an external party of result information, including the voice anomaly detection result obtained by the above-mentioned detection step.

Images