Computer-implemented emergency handling procedure and emergency communications network

Description

[0001] The present invention relates to a computer-implemented method of processing an emergency incident and a corresponding emergency communication network.

[0002] A public-safety answering point, PSAP, is a call center where emergency calls initiated by any mobile or landline subscriber are handled. However, during the handling of a plurality of emergency calls received basically at the same time for an emergency incident, the PSAP may be subject to an increased load. Amongst others, this may be caused by receiving not only audio calls, but also emergency calls containing resource consuming video streams.

[0003] In this respect, a further problem may arise in that the calls may be initiated using different codecs, and therefore, the streaming requirements for these calls could overload the resources of the PSAP.

[0004] Document US2020 / 162880A1 relates to a system for validating and supplementing emergency call information by detecting that multiple signals, including emergency calls, relate to the same event, and by deriving enhanced location, context, severity, and truthfulness data. Document US 2011 / 069625A1 disclose a method for dynamically optimizing bandwidth usage during real-time communications, by adjusting codec parameters and prioritizing ongoing calls based on various criteria such as user profile, call importance, and network conditions.

[0005] Therefore, the present invention is based on the object to provide a resource-saving and efficient computer-implemented method of processing an emergency incident and a corresponding emergency communication network.

[0006] This object is solved according to the present invention as set out in the appended claims, by a method of processing an emergency incident having the features according to claim 1, and a corresponding emergency communication system having the features according to claim 4. Preferred embodiments of the invention are specified in the respective dependent claims.

[0007] Thus, according to the present invention, a computer-implemented method of processing an emergency incident reported to a PSAP by a plurality of callers is provided, the method comprising the steps of: receiving, at the PSAP, a video call for reporting an emergency incident at a specified location; checking, at the PSAP, if further video calls have been received from the same specified location within a predetermined time period, and if it is determined that there are further video calls that have been received at the PSAP for the same specified location within the predetermined time period, feeding at least a part of the video call and the further video calls to a ML classifier unit, identifying, at the ML classifier unit, if there are similarities between the video call and the further video calls, and if it is determined that there exists similarity between the video call and at least one further video call, determining which one of the video calls that have been determined to be similar has a video stream that uses less resources, in particular, less bandwidth.

[0008] According to the present invention, the video call and the further video calls use different audio and / or video codec.

[0009] According to the present invention, the video stream that has been determined to use less resources is transmitted to a plurality of call takers that handle the emergency incident at the PSAP, and is displayed at a display unit of each of the call takers together with a current video stream they are already handling,

[0010] According to the present invention, the method further comprises a step of providing, to each of the call takers that handle the emergency incident at the PSAP, a button for selecting a video stream to be used for further handling the emergency incident and receiving selection of at least one call taker.

[0011] According to the present invention, the method further comprises a step of presenting the selection of the at least one call taker that handles the emergency incident at the PSAP to other call takers at the PSAP that handle the emergency incident, and asking the call takers to confirm their selection.

[0012] According to a preferred embodiment of the invention, the method may further comprise a step of generating, after confirmation of the selection by one of the call takers to switch to the video stream that uses less resources, at least one re-INVITE message that aims at converting the video call to audio call, to the emergency callers of the plurality of the emergency callers whose video calls are handled by said call taker select to switch from video call to audio call, other than the emergency caller of said video call having the video stream that uses less resources.

[0013] According to the present invention, the method further comprises a step of converting the video calls handled by the call takers who confirmed the selection to switch to the video stream that uses less resources, other than the video call having the video stream that uses less resources, from video calls to audio calls.

[0014] According to the present invention, the method further comprises a step of replicating the video stream that uses less resources to the call takers who confirmed the selection to switch to the video stream that uses less resources.

[0015] Further, according to the present invention, an emergency communication system is provided comprising at least one PSAP for handling at least one emergency incident and a Machine Learning, ML, classifier unit, said system being configured to perform the computer-implemented method for handling an emergency incident.

[0016] According to yet another preferred embodiment of the invention, the ML classifier unit uses a stream replication technique for determining similarity between the video streams.

[0017] According to the present invention, if it is determined that there are several video calls reporting the same emergency incident, then the video stream that has the least bandwidth requirements is determined and presented to a call taker at the PSAP who handles the emergency incident so that he or she may select which one is more efficient to use. If the video stream that uses the least bandwidth is sufficient for handling the emergency incident, then the other video streams that relate to the same emergency incident may be switched to simple audio calls, thereby saving resources, in particular, in terms of bandwidth usage.

[0018] The invention and embodiments thereof will be described below in further detail in connection with the drawing. Fig. 1schematically shows a scenario at a PSAP receiving and processing an emergency incident according to prior art; Fig. 2schematically shows another scenario at a PSAP receiving and processing an emergency incident according to an embodiment of the invention; Fig. 3shows a flow chart of the procedure of the method of processing an emergency incident according to an embodiment of the invention; and Fig. 4shows an end-to-end scenario according to still another embodiment of the invention.

[0019] Fig. 1 schematically shows a scenario at a PSAP 1 receiving and processing an emergency incident according to prior art. In the scenario illustrated here, there are exemplarily shown four emergency callers 2, 2', 2", 2‴ who respectively report the same emergency event or emergency incident by means of video, the video streams 4, 4', 4", 4‴ being transmitted to the PSAP 1 and from there being distributed to respective call takers or agents 3, 3', 3", 3‴ within a predetermined period of time, namely, for a point of time T0 to a point of time T1. According to this scenario, every call or video stream 4, 4', 4", 4‴ received at the PSAP 1 is transmitted to an individual call taker 3, 3', 3" 3,"'. For example, the video-based emergency call initiated by the caller 2 is transmitted as emergency video stream em.inc1 indicated by reference numeral 4 via the PSAP 1 to the call taker 3 who will process the information received and handle the emergency call. Another call taker 3' will handle the emergency video stream2 em.inc1 indicated by reference numeral 4', a further call taker 3" will handle the emergency video stream 3 em.inc1 indicated by reference numeral 4", and yet a further call taker 3‴ will handle the video stream 4 em.inc1 indicated by reference numeral 4‴. As can be seen, a lot of resources are used within the time period T0 to T1. Specifically, a lot of bandwidth is used by the plurality of video streams, although for reporting the emergency incident, it would be sufficient to simply transmit only one video stream for handling it.

[0020] Fig. 2 schematically shows another scenario at a PSAP 1 receiving and processing an emergency incident according to an embodiment of the invention. Again, exemplarily, there are four single callers 2, 2', 2", 2‴ who respectively report the same emergency incident by using video so that four different video streams, namely, emergency video stream1 em.inc1 indicated by reference numeral 4, which comprises CodecK for audio and CodecL for video, emergency video stream2 em.inc1 indicated by reference numeral 4', which comprises CodecM for audio and CodecA for video, emergency video stream3 em.inc2 indicated by reference numeral 4", which comprises CodecO for audio and CodecS for video, and emergency video stream4 em.inc1 indicated by reference numeral 4‴, which comprises CodecW for audio and CodecU for video. All four video streams 4, 4', 4'', 4‴ are received at the PSAP 1 where they are first transmitted to a Machine Learning, ML, classifier 5. Here, a stream replication technique is applied the main features of which are basically known from prior art. However, according to the embodiment illustrated here, the difference compared to prior art techniques is that the streams are compared in order to track which ones of them relate to the same emergency incident. That is, assuming there have been received four active video-based emergency calls at the PSAP 1 as outlined above for the illustrated example, the ML classifier 5 will identify which streams refer on the same emergency incident. The call takers 3, 3', 3", 3‴ handling these video streams 4, 4', 4", 4", will be presented with a short clip of the most lightweight stream resources. "Lightweight" stream in this context means the most efficient stream in terms of resources, i. e., bandwidth. After that, the call takers or agents 3, 3', 3", 3‴ will be in position to select if this option is better compared to the video stream they are already watching on their respective monitor or display unit. If the new lightweight stream provides the same pieces of information as to the handling of the emergency incident, a selection may be made by the respective agent or call taker which video stream he or she would like to use for handling or further processing the emergency incident. This selection may also be presented to the other call takers or agents who handle the same emergency incident. After this, all of the call takers 3, 3', 3", 3‴ will be in a position to verify which video stream 4, 4', 4'', 4‴ they want to use for handling and further processing the emergency incident.

[0021] In the example depicted in Fig. 2, the ML classifier 5, after tracking similarity between, for example, the video streams 4, 4', and 4‴, wherein the most efficient stream in terms of resources (bandwidth) is determined and forwarded to the call takers 3, 3', 3‴. In this case, the most efficient stream is the one that uses codecA for the video stream. The respective audio streams remain intact.

[0022] Fig. 3 shows a flow chart of the procedure of the method of processing an emergency incident according to an embodiment of the invention. Here, in the first step S1, a new emergency call comprising video data arrives at the PSAP 1. In the second step S2, the call taker responds to the emergency call. Then, in step S3, a check is carried out for identifying whether other video calls have been received at the PSAP 1 within the predetermined time period. If other calls have been received at the PSAP 1, then, in step S4, parts of the Real Time Protocol, RTP, stream of the current call comprising the video data and the other calls are fed to the ML classifier, 5 assuming that all calls refer to the same vicinity, i.e. geolocation. In the next step S5, a check is performed for determining, if there is at least one more similar emergency call. If at least one similar emergency call exists in the PSAP 1, in step S6, a check is performed for identifying which one of the calls received has the most efficient video stream in terms of bandwidth usage ("lightweight" codec). In step S7, the call takers of all similar calls (i.e., handling the same emergency incident) are presented with parts of the RTP streams of the available options. That is, they are presented together with the current stream, the one which corresponds to the active call they are already handling, and the most "lightweight" one. In step S8, a button is provided that enables the call takers to select which stream it is better for them in order to continue handling the emergency incident. In step S9, the call takers select the stream that fits better for the handling of the emergency incident. In step S10, the choices of each call taker as to the stream they have selected, is presented to the rest of the call takers. This is done with the aim to help the call takers to select the correct stream by also taking the selection of the other agents or call takers located at the PSAP 1 into account. After the choices are presented to the call takers, they are asked again, if they want to maintain their selection in step S11. Finally, in step S12, each one of the call takers who selected to switch over to the most lightweight video stream receives a re-negotiation message. The option for re-negotiation relates to the process of converting the video-based streams to simple audio streams. At the same time, the PSAP 1 replicates the video stream of the selected stream to the rest of the call takers (Step S13). Accordingly, the call takers are in position to switch back on the original video stream at any point of time during the handling and processing of the emergency call.

[0023] It is noted that the comparison between the RTP stream and the videos or images may be done using ML-based frameworks. Such comparisons can be accomplished by implementing, for example, the following frameworks: Karpathy, A., Toderici, G. Shetty, S. Leung, T., Sukthankar, R. & Fei-Fei L.: Large scale video classification with convolutional neural networks, in: Proceedings of the IEEE conference on Computer Vision and Pattern Recognition, pages 1725 - 1732.

[0024] Fig. 4 shows an end-to-end scenario according to still another embodiment of the invention. Here, it is assumed that four emergency callers 2, 2', 2'', 2‴ establish an emergency call received at the PSAP 1, using different codecs, for two different emergency incidents. The proposed method for handling and processing an emergency incident according to the embodiment detects the video stream with the minimum requirements and proposes this stream to the call takers 3, 3', 3", 3‴. The latter confirm and select the most efficient stream. In this example, the second call is the most efficient one.

[0025] Here, in the example described, the first three calls are already active, and the method is applied to the fourth call. That is, in STEP1, a new video call arrives on the PSAP 1. In STEP2, the call taker 3‴ responds to the call. In STEP3, a check is performed whether other active emergency video calls have been received at the PSAP 1. The same vicinity (geolocation) is considered in order to filter the different calls. Thus, only the calls from the same vicinity (geolocation, cell ID) are considered. In STEP4, parts of the video streams from the active calls and the examined call are fed into an ML classifier 5, in order to identify if there is a similarity between the calls. In STEP5, a match is found for the fourth call number initiated by the caller 2‴, and the calls initiated by the callers 2, 2'. These calls refer on the first emergency incident. Thus, the ML classifier 5 returns a positive result regarding the similarity of the first, second, and fourth calls. In STEP6, a check is performed in order to identify which ones of these calls require the minimum resources. In the current scenario, the most efficient stream corresponds to the second call made by the caller 2'. After this, the stream representing the most efficient call is presented to the call takers who handle the same incident. In this scenario, the stream of the second call made by the caller 2' will be presented to callTakerl indicated by reference numeral 3, callTaker2 indicated by reference numeral 3'; and the callTaker4 indicated by reference numeral 3‴. In STEP7, an option is displayed on the monitor of the previously mentioned call takers 3, 3', 3"': "Which one is the best stream for you?" In STEP8, the different selections are returned to the PSAP 1. In STEP9, the selection of every call taker is presented to the rest of the call takers who handle the same emergency incident.

[0026] For example, that callTaker1, is presented with the current stream of the active call, and the stream of call2 (only the video part of the most efficient stream). CallTaker2, is presented with the current stream of the active call. Additionally, an indication is appeared on her monitor which indicates that currently this is the most efficient stream among the similar calls in the PSAP element. CallTaker4 is presented with the current stream of the active video call which also handles the stream of call2. Having the previous parameters in mind, it is assumed that in STEP8, the callTakerl indicated by reference numeral 3 selects the stream of call2 initiated by the caller 2'. The callTaker2 indicated by reference numeral 3 maintains the same video stream, and the callTaker4 indicated by reference numeral 3‴ also selects the stream of call2 initiated by the caller 2'. Thereafter, in STEP10, the various selections are presented to every call taker 3, 3', 3", 3"'. For example, callTaker1 indicated by reference numeral 3 is presented with the selection of callTaker2 indicated by reference numeral 3' (i.e., preserves the stream of call2 imitated by the caller 2') and callTaker4 indicated by reference numeral 3‴ (i.e., selects the stream of call2 imitated by the caller 2'). This is done in order to help the call takers identify which are the different selections with regards to the specific emergency incident. Thus, in STEP11, a button is displayed at the respective display (not shown) at each call taker 3, 3', 3", 3‴ in order to verify his or her selection. In STEP12, callTaker1 indicated by reference numeral 3 and callTaker4 indicated by reference numeral 3‴ verify that they want to switch to the stream that is generated from call2 initiated by the caller 2'.

[0027] The previous selection generates re-INVITE messages for the emergency callers 2 and 2‴ that aim at switching the video-calls to simple audio calls. If the re-negotiation completes successfully, then the video stream is replicated on the callTaker1 indicated by reference numeral 3 and callTaker4 indicated by reference numeral 3"'.

[0028] The stream between the emergency caller2 indicated by reference numeral 2' and the callTaker2 indicated by reference numeral 3' is remained intact. On the contrary, the streams between the emergency caller1 indicated by reference numeral 2 and callTaker1 indicated by reference numeral 3 and the emergency caller4 indicated by reference numeral 2‴ and callTaker4 indicated by reference numeral 3‴ are converted to simple audio streams. The call takers of the first call initiated by the caller 2 and the fourth call initiated by the caller 2‴ are presented with the same video stream that is also presented to callTaker2 indicated by reference numeral 2'.Reference numerals

[0029] 1PSAP 2, 2', 2", 2‴emergency caller 3, 3', 3", 3‴call taker or agent 4, 4', 4", 4‴video stream 5Machine Learning Classifier

Claims

1. Computer-implemented method of processing an emergency incident reported to a Public-Safety Answering Point, PSAP, (1) by a plurality of callers (2, 2', 2", 2"'), the method comprising the steps of: - receiving, at the PSAP (1), a video call (4, 4', 4", 4‴) for reporting an emergency incident at a specified location; - checking, at the PSAP (1), if further video calls (4, 4', 4", 4‴) have been received from the same specified location within a predetermined time period, and if it is determined that there are further video calls (4, 4', 4", 4‴) that have been received at the PSAP (1) for the same specified location within the predetermined time period, feeding at least a part of the video call and the further video calls to a Machine Learning, ML, classifier unit (5), characterized in that the method further comprises - identifying, at the ML classifier unit (5), if there are similarities between the video call and the further video calls, the video call and the further video calls using different audio and / or video codec, and if it is determined that there exists similarity between the video call and at least one further video call, - determining which one of the video calls that have been determined to be similar has a video stream that uses less resources, in particular, less bandwidth, - transmitting the video stream that has been determined to use less resources to a plurality of call takers (3, 3', 3", 3‴) that handle the emergency incident at the PSAP (1) and displaying at a display unit of each of the call takers (3, 3', 3", 3‴) the video stream that uses less resources together with a current video stream they are already handling, - providing, to each of the call takers (3, 3', 3", 3‴), a button for selecting a video stream to be used for further handling the emergency incident and receiving selection of at least one call taker (3, 3', 3'', 3‴), - presenting the selection of the at least one call taker (3, 3', 3", 3‴) to other call takers (3, 3', 3", 3‴) at the PSAP (1) that handle the emergency incident, and asking the call takers to confirm their selection, - converting the video calls handled by the call takers who confirmed the selection to switch to the video stream that uses less resources, other than said video call having the video stream that uses less resources, from video calls to audio calls, and - replicating the video stream that uses less resources to the call takers who confirmed the selection to switch to the video stream that uses less resources.

2. Computer-implemented method according to the preceding claim, wherein the method further comprises a step of generating, after confirmation of the selection by one of the call takers to switch to the video stream that uses less resources, at least one re-INVITE message that aims at converting the video call to audio call, to the emergency callers (2, 2', 2", 2‴) of the plurality of the emergency callers whose video calls are handled by said call taker, other than the emergency caller of said video call having the video stream that uses less resources.

3. Computer-implemented method according to any one of the preceding claims, wherein the ML classifier unit (5) uses a stream replication technique for determining similarity between the video streams.

4. Emergency communication system, comprising at least one Public-Safety Answering Point, PSAP, (1) and a Machine Learning, ML, classifier unit (5), said system being configured to perform the computer-implemented method according to any one of the preceding claims.