JP7349072B2 - Voice recognition system for elevators - Google Patents

Description

TECHNICAL FIELD The present invention relates to an elevator speech recognition system for recognizing the voices of elevator passengers.

Conventionally, some elevators are equipped with a voice recognition system that recognizes the voice of passengers.For example, the elevator disclosed in Patent Document 1 recognizes the voice collected by a microphone inside the car, and then outputs the recognized voice. The car is equipped with a voice recognition system that includes a voice recognition device that recognizes the situation inside the car based on the following information.

Japanese Patent Application Publication No. 2011-073819

By the way, the in-car microphone of the above-mentioned conventional voice recognition system receives not only the voices of the passengers targeted for voice recognition, but also the voices of passengers not targeted for voice recognition and environmental sounds. The device may not be able to correctly recognize speech.

SUMMARY OF THE INVENTION In view of these circumstances, it is an object of the present invention to provide a voice recognition system for elevators that can improve the accuracy of voice recognition.

The voice recognition system for elevators of the present invention includes:
a sound processing means that executes a process of extracting voice data of a passenger from sound data collected in the car and identifying the location of the source of the voice data;
an image processing means that executes processing to identify the position of the passenger based on image data captured inside the car;
A voice recognition means is provided for performing voice recognition on the voice data extracted by the sound processing means, in which the position of the source is the same as the position of the passenger identified by the image processing means.

The voice recognition system for elevators having the above configuration is configured such that the voice recognition means can perform voice recognition on voice data in which the correspondence between the position of the source of the voice of the passenger and the position of the passenger is the same. This prevents voice recognition from being performed on voice data different from the voice data to be subjected to voice recognition, thereby improving the accuracy of voice recognition.

The voice recognition system for elevators of the present invention includes:
The sound processing means may be configured to include a sound feature extraction means for extracting sound feature information indicating characteristics of the passenger who is the source of the sound based on the sound data.

In this way, since information indicating the characteristics of the passenger can be used, it is possible to determine whether or not the passenger is an object to be subjected to voice recognition, and to improve the accuracy of voice recognition.

The voice recognition system for elevators of the present invention includes:
The image processing means may be configured to include a subject feature extraction means for extracting subject feature information indicating characteristics of the passenger appearing in the image data.

Even in this case, since information indicating the characteristics of the passenger can be used, it is possible to determine whether or not the passenger is a target for voice recognition, and to improve the accuracy of voice recognition.

The voice recognition system for elevators of the present invention includes:
comprising behavior determination means for determining whether or not the passenger is a target for voice recognition based on the passenger's behavior reflected in the image data;
The voice recognition means is configured to generate the voice in which the position of the source is the same as the position of the passenger indicated by the image data determined by the behavior determination means to be the passenger to whom the voice recognition is to be performed. The data may be configured to perform voice recognition.

In this way, it is possible to determine whether or not the passenger should perform voice recognition using the movement of the passenger, thereby increasing the accuracy of determining whether or not the passenger should perform voice recognition. I can do it.

The voice recognition system for elevators of the present invention includes:
comprising word determination means for determining whether or not the passenger is a target for voice recognition based on words extracted based on the voice data;
The voice recognition means may perform voice recognition on the voice data for which the word determination means has determined that the passenger is a passenger to whom voice recognition is to be performed.

In this way, it is possible to determine whether or not the object should be subjected to voice recognition based on the passenger's words, thereby increasing the accuracy of determining whether or not the object is to be subjected to voice recognition. I can do it.

As described above, the voice recognition system for elevators according to the present invention can achieve the excellent effect of improving the accuracy of voice recognition.

FIG. 1 is a block diagram schematically showing the configuration of a voice recognition system for elevators according to an embodiment of the present invention. FIG. 2 is an explanatory diagram of passenger information used in the elevator voice recognition system according to the embodiment. FIG. 3 is an explanatory diagram of image data used in the elevator voice recognition system according to the embodiment. FIG. 4 is a main flowchart of the voice recognition system for elevators according to the same embodiment. FIG. 5 is a sub-flowchart of the elevator speech recognition system according to the same embodiment, and is a sub-flowchart showing the flow of processing for creating speech-related information. FIG. 6 is a sub-flowchart of the voice recognition system for an elevator according to the same embodiment, and is a sub-flowchart showing the flow of processing for creating an image processing means. FIG. 7 is a sub-flowchart of the elevator voice recognition system according to the same embodiment, and is a sub-flowchart showing the flow of processing for creating passenger information. FIG. 8 is a sub-flowchart of the voice recognition system for an elevator according to the same embodiment, and is a sub-flowchart showing the flow of processing for determining whether or not a passenger is a contact person.

DESCRIPTION OF THE PREFERRED EMBODIMENTS A voice recognition system for elevators (hereinafter referred to as voice recognition system) according to an embodiment of the present invention will be described below with reference to the accompanying drawings.

The voice recognition system is a system configured to perform voice recognition on the voices of passengers in the car. Furthermore, in this embodiment, it is possible to identify a passenger who is trying to contact someone outside the car (referred to as a contact person in this embodiment) among the passengers in the car, and then to recognize the voice of this passenger. The speech recognition system will be explained by taking as an example the structure of the speech recognition system.

Note that a car that is subject to voice recognition in a voice recognition system includes, for example, as shown in FIG. Any device may be used as long as it is equipped with an external communication device T (see FIG. 3) for communicating with the outside.

The voice recognition system 1 of the present invention includes a sound processing means 2 that executes processing for creating voice-related information including voice data of each passenger based on sound data collected in the car; An image processing means 3 that executes processing for creating subject-related information including subject data in which the passenger is photographed for each passenger based on image data captured inside the car; Passenger information creation means 4 that creates passenger information by associating it with subject-related information; object selection means 5 that selects a passenger to perform voice recognition based on the passenger information; and object selection means 5. The vehicle includes a voice recognition means 6 that performs voice recognition on voice data of a passenger selected as a target for voice recognition.

The sound processing means 2 includes a sound data acquisition means 20 that acquires sound data, a sound creation means 21 that creates sound data for each passenger from the sound data acquired by the sound data acquisition means 20, and a sound created by the sound creation means 21. audio position deriving means 22 that derives the source of the sound (information indicating the position of the passenger, referred to as audio position information in this embodiment) based on the sound data obtained by the passenger; and the sound data extracted by the sound generating means 21. and a voice feature deriving means 23 for deriving information (referred to as voice feature information in this embodiment) indicating the characteristics of the passenger who is the source of the voice based on the voice characteristic information.

The sound data acquired by the sound data acquisition means 20 is the sound data collected by the sound collection device M installed in the car. Further, the sound data acquisition means 20 may acquire sound data directly from the sound collection device M, or may acquire sound data stored in the storage means after the sound collection device M has collected the sound data. (In other words, it may be configured to indirectly acquire sound data from the sound collection device M.

The voice creation means 21 is configured to perform a noise removal process that removes noise from sound data, and a voice extraction process that creates voice data for each passenger from the sound data from which noise has been removed in the noise removal process. has been done.

The noise that is removed from the sound data in the noise removal process is the sound that can be collected inside the car when there are no passengers, such as the sound of the car operating, announcements inside the car, and the external environment that enters from outside the car. It refers to sounds, etc.

Furthermore, the noise removal process acquires noise data created in advance, and extracts noise components from the sound data (sounds that are the same or approximately the same as the sound components included in the noise data) based on the noise data. components).

In the audio extraction process, audio data may be created by cutting out a portion corresponding to the passenger's voice from the sound data, or audio data may be created by separating the passenger's voice component from the sound data. may be configured. That is, the audio extraction process may be configured to create audio data containing the audio of one passenger.

The sound position deriving means 22 may be configured to derive sound source position information based on a plurality of sound data collected by a plurality of sound collecting devices M installed in the car, for example. In this case, the sound position deriving means 22 compares the plurality of sound data to derive the difference in sound propagation speed and the difference in sound pressure, and uses these information and information such as the installation position of each sound collection device. It is sufficient that the position of the source of the sound is derived based on the above.

Further, the audio position information may be information indicating the position of a predetermined area within the car (for example, a plurality of areas set by dividing the inside of the car in the front-rear direction and the left-right direction), The information may also be information indicating coordinates in the front-rear direction and left-right direction within the car.

The voice feature deriving means 23 performs a voice component derivation process that derives the characteristics of a voice component (passenger's voice component) based on the voice data, and indicates the characteristics of the passenger based on the voice component extracted in the voice component derivation process. The audio feature derivation process is configured to derive audio feature information.

The audio component derivation process may be configured to derive the passenger's audio component from the audio data, for example, by cepstrum analysis. The audio components derived in the audio component derivation process include, for example, the loudness, pitch, and timbre of the sound.

The audio feature deriving process may be configured to derive features related to the appearance of the passenger, such as gender and age group, based on the audio components extracted in the audio component deriving process.

Here, as shown in FIG. 2, in the sound processing means 2 of this embodiment, the sound source position information D11 extracted by the sound position deriving means 22 is associated with the sound data D10, and the sound feature deriving means 23 performs the sound feature deriving process. The voice feature information D12 derived in step 1 is associated with the voice data D10. Therefore, the audio related information D1 includes audio data D10, sound source position information D11, and audio feature information D12.

The image processing means 3 includes an image data acquisition means 30 that acquires image data, a subject extraction means 31 that extracts subject data for each passenger from the image data acquired by the image data acquisition means 30, and a subject extraction means 31 that extracts subject data for each passenger. The subject position deriving means 32 derives the position information of the passenger (hereinafter referred to as subject position information) based on the subject data obtained, and the subject extracting means 31 extracts the behavior of the subject (passenger) based on the subject data extracted. a behavior deriving means 33 for deriving the behavior; and a subject characteristic deriving means 34 for deriving information indicating the characteristics of the subject (passenger) (referred to as subject characteristic information in this embodiment) based on the subject data extracted by the subject extracting means 31. , has.

The image data acquired by the image data acquisition means 30 is an image captured by an imaging device C (for example, a camera, etc.) installed in the car. Further, the image data acquisition means 30 may acquire image data directly from the imaging device C, or may be configured to acquire image data that has been captured by the imaging device C and then stored in the storage device. (In other words, it may be configured to indirectly acquire image data from the imaging device C).

The subject extraction means 31 specifies an area in which the passenger is photographed within the image data. In this embodiment, as shown in FIG. The camera is configured to designate a divided area P1 in which is photographed, and use the divided area P1 as subject data.

In FIG. 3, a passenger who is a contact person is labeled with a symbol "H1", and a passenger who is not a contact person is labeled with a symbol "H2".

Note that the image data P may be a moving image or a still image. When the image data P is a still image, for example, a plurality of chronologically consecutive still images may be treated as one image data.

The subject position deriving means 32 derives position information of the passenger (information indicating the position of the passenger within the car) based on the subject data extracted by the subject extracting means 31.

The subject position deriving means 32 may be configured to derive the passenger position information based on the position of the subject data (divided area) P1 within the image data P, for example. In this case, each divided area of the image data is associated with a corresponding position in the car in advance, and the subject position deriving means 32 determines the position of the subject data (divided area) P extracted by the subject extracting means 31. It is sufficient if the information is configured to be position information.

As shown in FIG. 1, the behavior derivation means 33 performs a behavior derivation process for deriving the passenger's behavior, and determines whether the passenger should contact the outside of the car based on the passenger's behavior extracted in the behavior derivation process. A contact person determination process that determines whether the passenger is the intended contact person, and a contact person that indicates that the passenger is the contact person who is trying to contact someone outside the car based on the determination result of the contact person determination process. information, or non-contact person information indicating that the passenger is not a contact person attempting to contact someone outside the car, with the subject data.

The behavior derivation process of this embodiment is configured to determine whether or not the occupant is behaving, and if it is determined that the occupant is behaving, to derive the type of behavior of the occupant and the orientation of the occupant. has been done.

The behavior derivation process is configured, for example, to derive, as types of passenger behavior, speech actions that are recognized as the passenger making a speech, and non-speech actions that are recognized as the passenger not making a speech. That's fine.

Further, the behavior derivation process is configured to derive, for example, whether the passenger is facing the external communication device, or whether the passenger is facing a direction different from the external communication device. All you have to do is stay there.

The contact person determination process determines that the passenger is the contact person if the type of the passenger's behavior derived in the behavior derivation process is a speaking motion and the passenger is facing the external communication device, If the behavior derivation process determines that the passenger has no behavior, or the type of the passenger's behavior is derived to be non-speech, or the direction of the passenger is determined to be non-speech. If it is derived that the passenger is not the contact person, it is determined that the passenger is not the contact person.

As shown in FIG. 2, in the contact person information provision process, if the passenger is determined to be the contact person in the contact person determination process, the contact person information D22 is associated with the subject data D20. If it is determined that the person is not a contact person, the non-contact person information D23 is associated with the subject data D20.

The features of the passenger extracted by the subject feature extraction means are also features related to the passenger's appearance, such as gender and age group.

Here, in the image processing means 3 of this embodiment, the subject position information D21 extracted by the subject position deriving means 32 is associated with the subject data D20, and the subject characteristic information D4 derived by the subject characteristic deriving means 34 is also associated with the subject data D20. In addition, as described above, contact information D22 or non-contact information D23 is also associated with subject data D20. Therefore, the image-related information D2 includes subject data D20, subject position information D21, either contact information D22 or non-contact information D23, and subject characteristic information D24.

As shown in FIG. 1, the passenger information creation means 4 compares the audio location information associated with the audio data and the subject location information associated with the subject data, and compares the audio location information with the subject location information. If they indicate the same position, the camera is configured to create passenger information by associating the audio-related information with the subject-related information.

Note that the passenger information creation means 4 determines that the voice position information associated with the voice data and the subject position information associated with the subject data indicate the same position, and that the voice characteristic information indicates the passenger's position. It may be configured to create passenger information by associating the audio data with the subject data when the feature and the feature of the passenger indicated by the subject characteristic information correspond.

The target selection means 5 is configured to determine whether or not the passenger should be subjected to voice recognition based on the passenger's behavior and the words uttered by the passenger.

More specifically, the target selection means 5 includes a selection means 50 for selecting passenger information, and a passenger (in this embodiment, a contact person) who is to be subjected to voice recognition based on the passenger's behavior information. The behavior determining means 51 determines whether or not the passenger is the passenger who is the target of speech recognition based on the extracted words. and a word determining means 52 for determining whether or not.

The behavior determining means 51 is configured to determine that the passenger is not a passenger (contact person) to be subjected to voice recognition when non-contact person information is associated with the subject data.

The word determination means 52 determines whether or not the words extracted from the audio data include words for determination that are set in advance to determine whether or not the passenger is a contact person. If the words extracted from the voice data include the judgment word, the passenger is determined to be the contact person, and if the words extracted from the audio data do not include the judgment word, the passenger is not the contact person. is configured to make a determination. The word for judgment is, for example, a word indicating an abnormality of the car.

In a state where the contact person is not specified by the behavior judgment means 51 and the word judgment means 52, the voice recognition means 6 performs voice recognition of the voice data included in the passenger information selected by the selection means 50, and determines whether the contact person is In the specified state, the system is configured to perform voice recognition of the voice data included in the contact person's passenger information.

The configuration of the speech recognition system 1 according to this embodiment is as described above. Next, the operation of the speech recognition system 1 will be explained.

As shown in FIG. 4, in the voice recognition system 1, the sound processing means 2 creates voice-related information (S1), the image processing means 3 creates subject-related information (S2), and creates passenger information. (If the voice position information and the subject position information match) (Yes in S3), the passenger information creation means 4 creates passenger information based on the voice related information and the subject related information (S4) .

Next, it is determined whether the passenger information selected by the target selection means 505 belongs to the contact person (S5), and if the target selection means 505 determines that the passenger information belongs to the contact person ( If YES in S6), the voice recognition means 6 performs voice recognition on the voice data (S7).

Note that the voice recognition system of the present embodiment can be used after the voice recognition means 6 completes voice recognition of the voice data, or when passenger information is not created (No in S3), or when the target selection means 505 selects a passenger. If it is determined that the information does not belong to the contact person (No in S6), it is determined whether to end the process (Yes in S8) or continue the process (No in S8).

Further, although FIG. 4 shows the flow of processing by the image processing means 3 after the flow of processing by the sound processing means 2, the flow of processing by the sound processing means 2 is shown before the processing by the passenger information creation means 4 is executed. If the processing and the processing by the image processing means 3 can be completed, the processing by the sound processing means 2 may be executed after the processing by the image processing means 3, or the processing by the sound processing means 2 and the processing by the image processing means 3 may be executed. The processing may be performed in parallel.

As shown in FIG. 5, the sound processing means 2 repeats the process until the sound data acquisition means 20 acquires the sound data (No at S10), and when the sound data acquisition means 20 acquires the sound data (No at S10). Yes), the sound creation means 21 creates sound data for each passenger from the sound data acquired by the sound data acquisition means 20 (S11), and based on the sound data created by the sound creation means 21, the sound position derivation means 22 derives sound source position information (S12), and further, based on the sound data created by the sound creation means 21, the sound feature derivation means 23 derives sound feature information indicating the characteristics of the passenger (S13).

Then, the derived sound source position information and audio feature information are associated with the audio data, thereby creating audio related information (S14).

As shown in FIG. 6, the image processing means 3 repeats the process until the image data acquisition means 30 acquires the image data (No in S20), and when the image data acquisition means 30 acquires the image data (No in S20). Yes), the subject extraction means 31 extracts subject data for each passenger from the image data acquired by the image data acquisition means 30 (S21), and the subject position derivation means 32 extracts subject data for each passenger from the image data acquired by the image data acquisition means 30 (S21). The subject position information is derived (S22), and the behavior deriving unit 33 derives the behavior of the subject (passenger) based on the subject data extracted by the subject extracting unit 31, and based on the behavior, contact information or contact information is added to the subject data. Either one of the non-contact person information is associated (S23), and the subject feature deriving unit 34 derives subject characteristic information of the subject (passenger) based on the subject data extracted by the subject extracting unit 31 (S24).

Then, image-related information is created by associating the subject position information and subject characteristic information with the subject data to which either contact information or non-contact information is associated (S25).

As shown in FIG. 7, the passenger information creation means 4 compares the voice position information associated with the voice data and the subject position information associated with the subject data (S30), and compares the voice position information with the subject position information. The process of comparing whether the location information and the subject location information are the same is repeated (No in S31), and if the audio location information and the subject location information indicate the same location (Yes in S31), the audio related information and the subject related information are passenger information is created by associating them with each other (S32).

As shown in FIG. 8, the target selection means 505 selects passenger information when the selection means 50 selects the passenger information (S50), the behavior determination means 51 determines that the passenger is not the contact person (No in S51), and the word determination means If it is determined by 52 that the passenger is not the contact person (No in S52), a determination result indicating that the passenger is not the contact person is output (S53), and the word determination means 52 determines that the passenger is not the contact person. If so (No in S52), a determination result indicating that the passenger is not the contact person is output (S53).

On the other hand, if the behavior determining means 51 determines that the passenger is the contact person (Yes in S51) or the word determining means 52 determines that the passenger is the contact person (Yes in S52), boarding is prohibited. outputs a determination result indicating that the passenger is the contact person (S53); if the word determining means 52 determines that the passenger is not the contact person (No in S52), outputs a determination result indicating that the passenger is the contact person; The determination result is output (S54).

Then, as shown in FIG. 4, if the determination result is output that the passenger is the contact person (Yes in S6), voice recognition is performed by the voice recognition means 6.

Then, voice recognition is performed by the voice recognition means 6.

As described above, according to the voice recognition system 1 of the present embodiment, the voice recognition means recognizes the voice data in which the correspondence between the position of the source of the voice of the passenger and the position of the passenger's voice matches. Since the configuration is such that recognition can be performed, it is possible to prevent voice recognition from being performed on voice data different from the voice data to be subjected to voice recognition.

In this manner, the speech recognition system 1 of the present embodiment is capable of improving the accuracy of speech recognition by appropriately selecting the target of speech recognition.

In addition, since it is possible to use information indicating the characteristics of the passenger who uttered the voice, such as the voice feature information included in the voice related information and the subject feature information included in the subject feature information, voice recognition can be performed by using these information. It is also possible to improve the accuracy of voice recognition and the determination of whether or not the target should be subjected to voice recognition.

In particular, the voice feature information and subject feature information of this embodiment are features related to the passenger's appearance, so by using these information, it is possible to determine whether or not the subject should be subjected to voice recognition, Easy to improve recognition accuracy.

Note that, as described above, the passenger information creation means 4 determines that the voice position information associated with the voice data and the subject position information associated with the subject data indicate the same position, and the voice characteristic information is If the vehicle is configured to create passenger information by associating voice data and subject data when the passenger characteristics indicated by the subject characteristic information correspond to the passenger characteristics indicated by the subject characteristic information, the contact person's Identification errors can be suppressed.

Furthermore, in the voice recognition system 1 of the present embodiment, in addition to the correspondence between the position of the voice source (voice position information of voice-related information) and the position of the passenger (subject position information of subject-related information), Since the system is configured to determine whether the passenger is the contact person based on the passenger's behavior, the accuracy of selecting the target for voice recognition is improved.

In addition, the voice recognition system 1 of the present embodiment is configured to determine whether or not the passenger is the contact person based on the words uttered by the passenger. be able to choose flexibly.

Note that the voice recognition system for elevators according to the present invention is not limited to the above-described embodiments, and it goes without saying that various changes can be made without departing from the gist of the present invention.

Although not specifically mentioned in the above embodiment, the external communication device T is, for example, an intercom, a mobile information terminal (for example, a smart phone), or the like.

Although the sound position deriving means 22 of the above embodiment is configured to use a plurality of sound collection devices M, it may be configured to use one sound collection device M, for example. However, using a plurality of sound collecting devices M increases the accuracy of deriving the position of the sound source.

Although not specifically mentioned in the above embodiment, when repeatedly performing voice recognition by the voice recognition means 6, the sound processing means 2 for identifying the contact person, the image processing means 3, the passenger information creation means 4, the target After performing the series of processing by the selection means 5 again, the voice recognition means 6 may perform voice recognition, or the information already derived when identifying the contact person (for example, voice characteristic information, subject characteristic information, etc.) may be used. ) may be used to perform voice recognition by the voice recognition means 6 without performing a series of processes for identifying the contact person.

In the above embodiment, the voice recognition system 1 has been described using as an example a configuration in which a contact person is identified among the passengers in the car, and then the voice of this contact person can be recognized. However, the configuration is not limited to this. The voice recognition system 1 may be configured to, for example, identify a type of passenger other than the contact person and then recognize the voice of this passenger.

Although not specifically mentioned in the above embodiment, the voice recognition system 1 includes, for example, a storage device for storing information and a microcomputer for control, and the storage device and the microcomputer for control are installed in the car. It may be configured as follows. In this case, the storage device stores image data of the inside of the car, information indicating words (keywords) for judgment used by the word judgment means 52, etc. , the processing by the passenger information creation means 4, the object selection means 5, and the voice recognition means 6 is executed by a control microcomputer. Note that the information stored in the storage device may be in a database format, for example.

In the above embodiment, the sound collection device M is configured with the external communication device T, but the external communication device T collects sounds inside the car, such as an intercom. Any device may be used as long as it is equipped with a sound collection unit for generating sound and an output unit for outputting sound (more specifically, announcements to passengers, voices for interacting with passengers, etc.) inside the car.

Further, the sound collection device M can be configured by a device other than the external communication device T, and for example, it may be configured by a microphone installed in the car instead of the external communication device T. In this case, if a speaker is installed in the car and used as a substitute for the output section of the external communication device T, the external communication device T is not required and the sound collection function can be configured without duplication.

Although not specifically mentioned in the above embodiment, the voice recognition system 1 determines the situation inside the car based on sound data collected inside the car and image data captured inside the car. The system may be configured to take necessary actions (for example, control the operation of the elevator, ask questions to passengers, etc.) accordingly. In this way, it becomes possible to correctly determine what action to take based on the determined situation inside the car.

DESCRIPTION OF SYMBOLS 1... Voice recognition system, 2... Sound processing means, 3... Image processing means, 4... Passenger information creation means, 5... Target selection means, 6... Voice recognition means, 20... Sound data acquisition means, 21... Voice creation means , 22... Sound position deriving means, 23... Sound feature deriving means, 30... Image data acquisition means, 31... Subject extracting means, 32... Subject position deriving means, 33... Behavior deriving means, 34... Subject feature deriving means, 50... Selection means, 51... Behavior determination means, 52... Word determination means, 505... Target selection means, C... Imaging device, D1... Audio related information, D10... Audio data, D11... Sound source position information, D12... Audio feature information, D2 ...Image related information, D20...Subject data, D21...Subject position information, D22...Contact information, D23...Non-contact information, D24...Subject characteristic information, D4...Subject characteristic information, M...Sound collection device, P...Image Data, P1...Divided area, T...External communication device

Claims (5)

A sound collection device installed inside the car, which is composed of a sound collection unit that collects sounds inside the car, and an external communication device that includes an output unit that outputs sound for dialogue with passengers inside the car. a sound processing means for extracting voice data of the passenger from the sound data collected by the sound collection device and performing a process of specifying the location of the source of the voice data;
an image processing means that executes processing to identify the position of the passenger based on image data captured inside the car;
Among the voice data extracted by the sound processing means, the location of the source is the same as the position of the passenger identified by the image processing means, and the sound is transmitted by a contact person who is trying to contact an external party. voice recognition means for performing voice recognition on the object ,
Voice recognition system for elevators. The sound processing means includes a sound feature extraction means for extracting sound feature information associated with the sound data, the sound feature information indicating the characteristics of the passenger who is the source of the sound based on the sound data.
The voice recognition system for an elevator according to claim 1. The image processing means includes a subject characteristic extraction means for extracting subject characteristic information associated with subject data, which is subject characteristic information indicating the characteristics of the passenger reflected in the image data.
A voice recognition system for an elevator according to claim 1 or 2. comprising behavior determination means for determining whether or not the passenger is a target for voice recognition based on the passenger's behavior reflected in the image data;
The voice recognition means is configured to generate the voice in which the position of the source is the same as the position of the passenger indicated by the image data, which is determined by the behavior determination means to be the passenger to be subjected to voice recognition. perform voice recognition on data,
A voice recognition system for an elevator according to any one of claims 1 to 3. comprising word determination means for determining whether or not the passenger is a target for voice recognition based on words extracted based on the voice data;
The voice recognition means performs voice recognition on the voice data determined by the word determination means to be a passenger to be subjected to voice recognition.
A voice recognition system for an elevator according to any one of claims 1 to 4.