EP3871944A1 - Method for documentation of a voice-based communication - Google Patents

Abstract

In one embodiment, the method is used to document voice-based communication and comprises the steps: - providing a transmitter-side and / or a receiver-side voice recognition device (4), - speaking and transmitting a message (M) via a communication channel (3) from a transmitter (A ) to a receiver (B) and digitizing at the sender and / or receiver side and storage of the message (M) by means of the speech recognition device (4), and speaking of a confirmation (C) of the message (M) by the receiver (B) and sender-side and / or digitization at the receiver end and storage of the confirmation (C) by means of the speech recognition device (4).

Description

A method for documenting speech-based communication is specified. In addition, a route component, a vehicle component, a speech recognition device and a system for such a method are specified.

The pamphlet WO 2019/243181 A1 relates to a method and a device for controlling a rail vehicle by means of voice messages.

One task to be solved is to provide a method with which speech-based communication can be efficiently documented.

This object is achieved, inter alia, by a method, by a route component, by a speech recognition device and by a system with the features of the independent claims. Preferred developments are the subject of the dependent claims.

The method described here enables automatic speech recognition and digitization of voice radio, as well as display of transmitted messages and securing of communication via a return channel, in particular when operating rail vehicles. In this way, security-relevant information can be transmitted by voice radio and handwritten notation of messages and / or confirmations is no longer necessary.

• Bereitstellen einer senderseitigen und/oder einer empfängerseitigen Spracherkennungseinrichtung,
• Sprechen und Übermitteln einer Nachricht über einen Kommunikationskanal von einem Sender an einen Empfänger und senderseitiges und/oder empfängerseitiges Digitalisieren sowie Speichern der Nachricht mittels der Spracherkennungseinrichtung, und
• Sprechen einer Bestätigung der Nachricht durch den Empfänger und senderseitiges und/oder empfängerseitiges Digitalisieren sowie Speichern der Bestätigung mittels der Spracherkennungseinrichtung.

In at least one embodiment, the method is used to document voice-based communication and comprises the following steps, in particular in the specified order:

• Providing a transmitter-side and / or a receiver-side speech recognition device,
• Speaking and transmitting a message via a communication channel from a transmitter to a receiver and digitizing at the transmitter and / or receiver side and storing the message by means of the voice recognition device, and
• Speaking of a confirmation of the message by the recipient and digitizing at the sender and / or recipient side and storage of the confirmation by means of the speech recognition device.

1. 1. Der Fahrer und der Fahrdienstleiter kontaktieren einander, normalerweise über Telefon mittels des konventionellen Zugfunks.
2. 2. Der Fahrdienstleiter diktiert die Anweisung oder den Befehl dem Fahrer.
3. 3. Der Fahrer füllt dann ein Formular aus gemäß den Instruktionen des Fahrdienstleiters.
4. 4. Aus Sicherheitsgründen muss der Fahrer die Informationen, die er in das Formular eingetragen hat, an den Fahrdienstleiter berichten.
5. 5. Nachdem der Fahrer die Anweisungen vollständig ausgeführt hat, muss er dies in dem Formular vermerken, so dass später eine Rekonstruktion möglich ist, um speziell im Falle eines Unfalls festzustellen, welche Aktionen hätten vorgenommen werden sollen oder vorzunehmen waren. Insbesondere nach dem Ausstreichen von Befehlen aus dem Formular sollen ausgestrichenen Befehle weiterhin lesbar bleiben.
6. 6. Der Fahrdienstleiter füllt ein separates Formular aus, das zu Beweiszwecken in der Betriebszentrale verbleibt.

If a signal transmission intended for a certain operating situation is not possible or deviates from the normal operating mode, especially of a railway, it is helpful to be able to transmit an instruction or an order from a dispatcher to a driver in some other way. The verbal transmission of instructions or commands is also referred to as "Travel Permitted on Command". A common application for this is driving into a hazard area without adequate trackside and / or vehicle-specific signals, for example in shunting operations or in a disturbed operation. In such a case, the process is as follows:

1. 1. The driver and dispatcher contact each other, usually by telephone using conventional train radio.
2. 2. The dispatcher dictates the instruction or command to the driver.
3. 3. The driver then fills out a form as instructed by the dispatcher.
4. 4. For safety reasons, the driver must report the information that he has entered in the form to the dispatcher.
5. 5. After the driver has carried out the instructions in full, he must note this in the form so that a reconstruction is possible later in order to determine, especially in the event of an accident, which actions should or should have been carried out. In particular, after commands have been crossed out from the form, crossed out commands should remain legible.
6. 6. The dispatcher fills out a separate form that remains in the operations center for evidence purposes.

The entire process, especially the documentation and confirmation of the instructions, is time-consuming and error-prone. Not only, but mainly because of the high level of human involvement. Another complicating factor is that the voice channel can be noisy. This enables both a simplification of this process and a relief for people through partial or complete automation of the process steps specified above, in particular through the use of artificial intelligence, or AI for short, for speech recognition and digitization of speech information and for generating and Processing a digital image.

The method described here enables language communication to be supported and secured in critical systems and, in particular, in railway operations. Thanks to the automatic speech recognition, the spoken message can be sent in digitized form in order to ensure greater reliability and security of communication. Digitizing and sending the message along relieve the recipient, as he does not have to memorize all the details of the information. Ultimately, the results of speech recognition can be used in business situations by linguistically controlling a device.

• transaktionsbasierte Sicherung der Kommunikation in einem für die Sprechenden transparenten Rücksendekanal,
• Spracherkennung und Digitalisierung von Sprachfunk und Anzeige auf einen HMI (Human Machine Interface),
• zuverlässige Spracherkennung und -verarbeitung sowie Unterdrückung negativer Auswirkungen akustischer Störungen.

This is made possible in particular by:

• transaction-based security of communication in a return channel that is transparent for the speaking,
• Speech recognition and digitization of voice radio and display on an HMI (Human Machine Interface),
• reliable speech recognition and processing as well as suppression of negative effects of acoustic disturbances.

The procedure described here can be used on all railway lines. By increasing the reliability of existing and by introducing new technical barriers, in the sense of technical safety, a simplification of the application of the procedure for drivers and dispatchers is achieved. As a direct consequence of this, faster and more efficient operation can be guaranteed. In particular, the method can serve as an operational fallback level for a disturbed ETCS level 3 operation, so that if the ETCS infrastructure fails, the train radio can be used to at least partially maintain operation.

According to at least one embodiment, the sender of the message is a vehicle component, in particular in a rail vehicle such as a train, and the receiver is a route component, or vice versa. The messages and confirmations are thus preferably exchanged between a driver, English driver, and a dispatcher, English dispatcher.

According to at least one embodiment, the speech recognition device compares whether the message and the confirmation have the same content. This means that the content of the message and the associated confirmation can be checked by means of the speech recognition device.

According to at least one embodiment, the speech recognition device issues a warning if a content-related agreement between the message and the confirmation is uncertain. This ensures that an automated content check of the messages and confirmations does not lead to any security risks.

According to at least one embodiment, the message and the confirmation are digitized in real time and localized at the sender and / or recipient. This means that the speech recognition device does not need any access to external computers or servers. This means that speech recognition can be carried out independently of external components.

According to at least one embodiment, a structure and a content of the message and of the confirmation are regulated. In particular, there is regulation by a set of rules of the operator of the corresponding railway line. In other words, due to the specific speaking situation, certain terms and sequences of the information contained are given. As a result, the speech recognition is fixed and / or limited to a limited vocabulary and to a certain regulated structure. This makes speech recognition easier and increases the level of security when checking the correspondence between the content of the message and the associated confirmation. For example, the vocabulary required for speech recognition is 1000 words or less or 500 words or less.

According to at least one embodiment, the message contains a command to be executed by the recipient. After executing the command contained in the message, the recipient confirms this to the sender with a spoken message. This message is digitized at the transmitter end and / or at the receiver end by means of the speech recognition device and preferably also stored. This can be used to automatically document whether the transmitted commands are being carried out. Optionally, a warning or a note can be generated automatically if no notification is made or if the notification is too long in coming.

According to at least one embodiment, a large number of messages, confirmations and, optionally, messages are transmitted via the communication channel. Depending on the operating mode of the rail vehicle in particular, the frequency of messages and associated confirmations can be in the region of one per minute. Without automation and speech recognition, the achievable frequency of messages and confirmations would be significantly reduced.

According to at least one embodiment, the speech recognition device creates a log of the messages, confirmations and, optionally, notifications. The log can be continuously updated and saved. It is also possible for the log to be automatically sent to a predetermined test center at predetermined intervals, for example at the end of a journey by the rail vehicle.

According to at least one embodiment, the communication channel is a train radio. In particular, the communication channel is a radio link in accordance with the GSM-R standard.

According to at least one embodiment, the speech recognition device is only present at the transmitter. Alternatively, the speech recognition device is only available at the recipient. This means that the method described here can also be carried out unilaterally, i.e. only on the receiver side or only on the transmitter side.

According to at least one embodiment, the speech recognition device is present both at the sender and at the receiver. In other words, both the transmitter and the receiver then preferably each have their own speech recognition device. That is, the message and the confirmation and optionally the message can be digitized and saved on the sender and receiver side.

In accordance with at least one embodiment, the method described here is carried out in the event of failure of a train control system as a bridging for train routing. The train control system is in particular ETCS.

In addition, a route component is specified. The route component, which is contained in a signal box, for example, is set up to carry out a method described here. Features of the method are therefore also disclosed for the route component and vice versa.

In addition, a vehicle component is specified, in particular a vehicle component in a rail vehicle, for example a train. The vehicle component is set up to carry out a method described here. Features of the method are therefore also disclosed for the vehicle component and in particular the rail vehicle, and vice versa.

In addition, a speech recognition device for a vehicle component, in particular in a rail vehicle, and / or for a route component is specified. The speech recognition device is set up to carry out a method according to one or more of the above-mentioned embodiments. Features of the method are therefore also disclosed for the speech recognition device and the vehicle component, and vice versa.

Furthermore, a system with a route component and with one or more vehicle components, in particular in rail vehicles, is specified. The system is set up for a method in accordance with one or more of the above-mentioned embodiments. Features of the system are therefore also disclosed for the method and vice versa.

The above-mentioned properties, features and advantages of the invention and the manner in which they are achieved are further explained by the following description of the exemplary embodiments of the invention in conjunction with the corresponding figures, wherein
the Figures 1 and 2
show schematic block diagrams of exemplary embodiments of the methods described here, and
the Figures 3 to 5
illustrate schematic representations of exemplary embodiments of the methods described here.

In the exemplary embodiments described here, both the transmission of the voice information and the transmission of the generated digital image and associated data take place, for example, using conventional train radio as communication channel 3. Alternatively or in addition to train radio, other communication channels 3 can also be used.

In general, train control takes place through an interaction of track-side and vehicle-side, in particular train-specific system components, which can originate from different technological epochs and from different manufacturers. The method described here can fall back on different implementation possibilities in different scenarios, for example on a vehicle-specific solution, a track-side specific solution and a solution for the implementation of automatic functions on both sides. This allows the method to be used in scenarios with different operating modes. The basic principle of "drive on command" and the associated interactions between the driver and the dispatcher can be retained.

For voice recognition, voice processing and transmission of a digital image of the voice message M, sent From the dispatcher as sender A to the driver as receiver B, as well as for the associated confirmation C, an information validation step is required for security reasons. There are two main implementation options for this, with a security feature such as a generated checksum of the digital image of the voice message M being sufficient to allow the message M to be compared with the confirmation C.

In Figure 1 A first variant is shown as an exemplary embodiment in which manual validation is carried out by the dispatcher, for example the transmitter A, in particular by means of a display device such as a human machine interface, or HMI for short. The HMI is located on a route component 1, for example in an interlocking where the dispatcher is also located. The spoken message M arrives at a voice recognition device 4. In the voice recognition device 4, voice recognition and voice processing of the message M take place. Furthermore, a security feature such as a checksum is preferably calculated in the voice recognition device 4. The associated information is exchanged, for example by means of train radio.

Thus, the driver, who is the recipient B of the message M and who is preferably in a rail vehicle 2, receives the voice message 31 himself as well as a steganographic digital feedback 33. The steganographic feedback 33 is thus also transmitted via the train radio as communication channel 3 but not perceived as such by driver B. Furthermore, the voice recognition device 4 returns digital feedback 32 to the dispatcher A, for example as a checksum. The digital feedback 32 can be processed and / or displayed by the HMI. In the same way, a driver machine interface, or DMI for short, can display the respective information to driver B.

The recipient B of the message M speaks a voice message 31 with an acknowledgment C from the voice recognition device 4 is processed in the same way as the underlying message M. Thus, a steganographic feedback 33 of the voice transmission 31 also reaches the recipient A and the digital feedback 32 reaches the driver B.

On the recipient side, a manual check can now be carried out by dispatcher A as to whether the content of the confirmation C corresponds to the message M.

In the procedure according to Figure 1 no handwritten transmission of the message M and / or the confirmation C is required on the recipient side and / or on the sender side in forms provided for this purpose. As an alternative to displaying the Figure 1 speech recognition takes place only on the receiver side or only on the sender side. In addition, in Figure 1 As in all other exemplary embodiments, the roles of receiver B and transmitter A can be reversed, so that in principle the driver can also act as transmitter A and the dispatcher as receiver B.

In the embodiment of Figure 2 it is shown that a validation unit 41 is also present, as is also possible in all other exemplary embodiments. A check is carried out in the validation unit 41 to determine whether the content of the confirmation C corresponds to the message M. It is possible for this check to take place on the basis of the digital feedback 32, that is to say, for example, on the basis of checksums.

The validation unit 41 can be a separate component or, deviating from the representation of FIG Figure 2 , be integrated in the speech recognition device 4.

The remarks on Figure 1 equally for Figure 2 .

Thus, by the method as in Figures 1 and 2 exemplifies an improvement in reliability and time savings when driving in accordance with a driver's license on command, ie "driving on command", achieved, since a simplified process flow is achieved compared to a purely manual, handwritten recording of the messages M and the confirmations C.

In particular, the procedures as in the Figures 1 or 2 illustrated can be used in connection with a train control system 6. In the case of an ETCS level 3 system 6, a track vacancy report is in principle no longer carried out on the track side, for example by means of an axle counter, but a position report from the ETCS vehicle equipment is sent to an RBC (Radio Block Center). When the current vehicle position is reported, the completion of the train is confirmed to the ETCS center so that route sections can be released accordingly. A safety-related check of the train integrity on the vehicle side is therefore necessary.

However, this very technology-centered approach requires high reliability and availability of the system components as well as continuous train-to-route communication, which requires a high-availability communication channel. If one of the system components is not available, in particular the communication channel between the train and the line, serious disruptions in operations are foreseen. This is almost inevitably the case, in particular, since national class B systems are usually no longer installed on new ETCS railway lines.

With the method described here, it is possible to use conventional train radio as a fallback level if ETCS components fail. The failure of the train control system 6 is in the Figures 1 and 2 symbolized by the fact that the train control system 6 is shown crossed out.

The train is then driven on the basis of commands that enable the journey. Such commands are particularly permitted from an independent ETCS stop signal to the next stop signal. The method described here can therefore be used to increase reliability, so that the method described here enables shadow operation or a fall-back level within the framework of ETCS level 3 operation. In this way, overall higher train speeds and an increased vehicle throughput can be achieved.

Several possible, preferred scenarios that may occur are explained schematically by way of example.

In a first step of the method of Figure 3 the driver B and the dispatcher A contact each other, the communication channel 3 preferably being a train radio. This process step takes place without additional assistance systems or automatic functions.

In the next step, dispatcher A dictates message M to driver B with a command O. Trackside system 1 uses voice recognition to generate and process digital, steganographic feedback 33, which is transmitted to driver B together with voice message 31. Furthermore, a checksum is generated by means of the speech recognition device 4 and displayed by means of the HMI. In this method step, there is therefore an assistance function, in particular through the speech recognition device 4.

On the vehicle side, the voice recognition device 4 on the vehicle 2 also performs voice recognition for generating and processing the voice message 31 with the command O. A digital image of the voice message 31 is preferably displayed on a DMI. An assistance function is thus also provided on the vehicle side.

In a subsequent method step, the vehicle is automatically stored S of the digital image of the command O for reporting and for legal reasons. This also has an assistance function.

In a further process step, driver B has to confirm the information he has received from dispatcher A for safety reasons. To this end, the vehicle-mounted system uses speech recognition to generate and process G a digital image and also to generate checksums for a report. For a validation, the checksum that was generated by the vehicle is transmitted to the route component 1. This corresponds to a further assistance function.

This is followed by a comparison T between the message M and the confirmation C.

In a first variant, the trackside system 1 uses the checksum of the digital image of the report in order to automatically validate whether the driver B has correctly confirmed the corresponding voice message 31 and thus the command O. For final information by the dispatcher A, the validated result is preferably displayed on the HMI of the system component 1. Again there is an assistance function.

In a second variant, the checksum of the digital image of the command O and the checksum of the digital image of the confirmation C are displayed so that the dispatcher can manually validate whether the driver B has correctly confirmed the associated message M. There is again an assistance function.

Finally, the command O contained in the message M is executed E by the driver B. In a message N, the driver B confirms the correct execution E of the command O. This is preferably done by means of the DMI, which again contains an assistance function.

In a further method step, on the track side, the digital image of the command O to the driver B and optionally the message M is preferably automatically stored S.

Furthermore, in Figure 3 illustrates that optionally, for example on the vehicle side, a protocol P is created on the message traffic. The protocol P can be generated automatically with the aid of the speech recognition device 4. Furthermore, the protocol P can be sent to a control instance, such as the route component 1, for documentation purposes. It is also possible, as an alternative or in addition, to create a further log on the track side, preferably automatically. Such protocols can also be used in all other exemplary embodiments.

In the procedure according to Figure 4 an automatic verification of the message M and the confirmation C of the driver B takes place on the vehicle side. In a first step, the driver B and the dispatcher A again establish contact with one another via the communication channel 3.

In the next step, dispatcher A dictates the message with command O to driver B, and the vehicle uses voice recognition device 4 for voice recognition and for generating and processing a digital image of message M. A digital image of message M is preferred on the DMI Driver B shown. Herein lies an assistance function.

In a next step, the vehicle-mounted system automatically saves the digital image of the message M and thus of the command O for reporting purposes and for legal reasons. This is preferably an automatic function.

For safety reasons, driver B then confirms the information that he has received from dispatcher A. The vehicle-mounted system uses speech recognition to generate and process a digital image of the confirmation C in order to automatically verify whether the driver B confirms the associated message M correctly. Again there is an assistance function.

The driver B then confirms the correct execution E of the command O displayed in particular on the DMI; this confirmation is preferably also made at the DMI. This provides an assistance function by means of the speech recognition device 4. In addition, the message N about the execution of the command O is automatically stored S for reporting purposes and for legal reasons. Optionally, a message N about the execution of the command O to the dispatcher A can be made.

Finally, dispatcher A fills out a separate form F for the above process, which remains in an operations center for evidence purposes. In contrast, the digital images of the message M, the confirmation C and the message N are automatically stored in the vehicle for reporting purposes and for legal reasons.

According to Figure 4 the speech recognition device 4 is therefore only present in at least one vehicle component on the rail vehicle 2 and not on the track side.

In the procedure according to Figure 5 if the speech recognition device 4 is only present on the route side of the route component 1. Again, a connection between dispatcher A and driver B is first established via communication channel 3.

The dispatcher B then dictates the message M with the command O to the driver B, with voice recognition being used on the track side for generating and processing a digital image of the message M and the digital image being used Dispatcher A is shown in particular by means of an HMI as an assistance function.

The driver B then fills out a form F which contains the command O corresponding to the message M from dispatcher B. In this step, there is neither an assistance function nor an automatic function on the vehicle side.

For safety reasons, driver B confirms the information he has received from dispatcher A. The trackside system uses speech recognition to generate and process a digital image of the confirmation C and preferably automatically validates whether the driver B has correctly confirmed the associated message M. This step is preferably an automatic function.

After driver B has fully executed command O, the command must be crossed out and the associated form filed so that it can be reconstructed later, especially in the event of an accident, which action should have been taken by driver B or which action was taken. After crossing out the command O in form F, the associated information must still be legible for legal reasons. Again, there is no automatic or assistance function on the vehicle side. Optionally, the message N about the execution of the command O is sent to the dispatcher A.

The trackside system automatically stores S the digital image of the message M, the confirmation C and, optionally, the message N.

The speech recognition and the speech processing by the at least one speech recognition device 4 are preferably checked for security and a risk assessment is preferably carried out, for example in accordance with EN 50129: 2018 and / or EN 50159: 2010.

Although the invention has been illustrated and described in detail on the basis of exemplary embodiments, the invention is not limited to the disclosed exemplary embodiments and the specific combinations of features explained therein. Further variations of the invention can be obtained by one skilled in the art without departing from the scope of the invention as claimed.

List of reference symbols

A.

Channel

B.

recipient

C.

confirmation

E.

Execution of the instruction

F.

form

G

Generation and processing of a digital image

M.

news

N

Message

O

command

P.

protocol

S.

storage

T

comparison

1

Route component

2

Rail vehicle

3

Communication channel

31

Voice transmission, in particular by means of train radio

32

digital feedback

33

steganographic digital feedback

4th

Speech recognition device

41

Validation unit

5

system

6th

Train control system

Claims (15)

Procedure for documenting voice-based communication with the following steps: - Provision of a transmitter-side and / or a receiver-side speech recognition device (4), - Speaking and transmitting a message (M) via a communication channel (3) from a transmitter (A) to a receiver (B) and digitizing at the transmitter and / or receiver side and storing the message (M) by means of the speech recognition device (4), and - Speaking of a confirmation (C) of the message (M) by the recipient (B) and digitizing at the transmitter and / or at the receiver and storage of the confirmation (C) by means of the speech recognition device (4). Method according to the preceding claim,
wherein the transmitter (A) is a vehicle component, in particular a rail vehicle (2), and the receiver (B) is a route component (1), or vice versa. Method according to one of the preceding claims,
wherein the speech recognition device (4) compares whether the message (M) and the confirmation (C) are authentic and have the same content,
wherein the speech recognition device (4) issues a warning if a content match between the message (M) and the confirmation (C) is uncertain. Method according to one of the preceding claims,
wherein the digitization of the message (M) and the confirmation (C) takes place in real time and localized at the sender (A) and / or at the recipient (B). Method according to one of the preceding claims,
wherein a structure and a content of the message (M) and the confirmation (C) are regulated, and
wherein the speech recognition device (4) for a limited Vocabulary is set up in accordance with the regulated structure and the regulated content of the message (M) and the confirmation (C). Method according to one of the preceding claims,
the message (M) containing a command (O) to be executed by the recipient (B),
wherein the receiver (B) confirms this to the transmitter (A) with a spoken message (N) after executing the command (O) and the message (N) is digitized and stored by the voice recognition device (4) at the transmitter and / or receiver side. Method according to one of the preceding claims,
wherein a large number of messages (M), confirmations (C) and optionally messages (N) are transmitted over the communication channel (3),
wherein the speech recognition device (4) creates a protocol (P) of the messages (M), confirmations (C) and optionally messages (N). Method according to one of the preceding claims,
wherein the communication channel (3) is a train radio. Method according to one of the preceding claims,
wherein the speech recognition device (4) is only present at the transmitter (A), or
wherein the speech recognition device (4) is only present at the receiver (B). Method according to one of Claims 1 to 8,
wherein the speech recognition device (4) is present both at the sender (A) and at the recipient (B), so that the message (M) and the confirmation (C) and optionally the message (N) are digitized and stored at the sender and receiver side. Method according to the immediately preceding claim, which is carried out in the event of failure of a train control system (6) as a bridge to the train control. Route component (1) which is set up to carry out a method according to Claim 9. Vehicle component, in particular a rail vehicle (2), which is set up to carry out a method according to claim 9. Speech recognition device (4) for a route component (1) or for a vehicle component, in particular a rail vehicle (2), the speech recognition device (4) being set up to carry out a method according to Claim 9. System (5) with a route component (1) and with a vehicle component, in particular a rail vehicle (2), the system being set up for a method according to one of Claims 10 or 11.

Images