JP2005080110A - Audio conference system, audio conference terminal, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an audio conference system by which a plurality of persons can participate in an audio conference from one point even without using a plurality of lines and identification of a speaker is easy. <P>SOLUTION: The audio conference terminal has a microphone which collects voice of the speaker and output voice information indicating the voice of the speaker, an identification means provided near the microphone and to output identification information based on owner information read from a recording medium in which the owner information for specifying the speaker is recorded, an identification information addition means to add identification information to the voice information and a transmission means to transmit the voice information to which the identification information is added. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to an audio conference system, and more particularly to an audio conference system in which a listener can easily distinguish speakers.

In recent years, with the globalization and speeding up of business, the importance of a communication conference system capable of conducting a conference in real time between geographically distant points is increasing.
When such a communication conference system is a so-called video conference system that involves communication of images (moving images) together with voice, it is easy for the listener to identify the speaker because the conference can be performed while watching the facial expression of the communication partner. In the case of an audio conference system without an image, it is difficult for the listener to identify the speaker.

In order to solve this problem, in a voice conference system using ISDN (Integrated Services Digital Network), (1) channel control means for allocating a communication channel for each speaker, and (2) voice for each speaker. An audio conference system that makes it easy to identify a speaker by changing an output position, changing an audio signal for each speaker, or displaying the name of the speaker has been proposed (for example, Patent Documents). 1).
JP-A-8-125738

  However, in the conventional technique using the ISDN line, it is necessary to prepare ISDN lines for the number of participants in the conference. That is, when a plurality of people participate in an audio conference from one point (for example, one conference room in the office), it is necessary to prepare ISDN lines for the number of people in the conference room, which is difficult to implement.

  The present invention has been made in view of the above circumstances, and a voice conference system in which a plurality of people can participate in a voice conference from one point without using a plurality of lines, and speaker identification is easy. The purpose is to provide.

In order to solve the above problems, the present invention collects a speaker's voice, outputs a voice information indicating the voice of the speaker, and an owner provided near the microphone to identify the speaker Identification means for outputting identification information based on the owner information read from the recording medium on which the information is recorded, identification information addition means for adding identification information to the voice information, and transmission of the voice information with the identification information added An audio conference terminal device having a transmitting means is provided.
According to this audio conference terminal device, the identification information for identifying the speaker who has emitted the audio is added to the audio information and transmitted, so that the receiving side can speak based on the identification information added to the audio information. Person can be identified.

Further, the present invention is based on a microphone that collects the voice of a speaker and outputs voice information indicating the voice of the speaker, a voice recognition unit that recognizes the voice of the speaker, and a recognition result of the voice recognition unit. There is provided an audio conference terminal device comprising: identification means for determining identification information; identification information adding means for adding identification information to voice information; and transmission means for transmitting voice information to which the identification information is added.
According to this voice conference terminal device, the identification information for identifying the speaker who has emitted the voice is included in the voice information based on the voice recognition result without using a recording medium in which the identification information of the speaker is recorded. Identification information for identifying the speaker is acquired and added to the voice information.

Further, the present invention provides a receiving means for receiving voice information to which identification information for identifying a speaker is added, a feature adding means for adding a feature to the voice indicated by the voice information based on the identification information, and a feature addition Provided is an audio conference terminal device having audio output means for outputting audio with features added by the means.
According to this voice conference terminal device, the voice information to which the identification information for identifying the speaker who has made the voice is added is received, and the voice characterized by each speaker is reproduced. Can be easily identified.

  Further, the present invention collects voice of a speaker and outputs voice information indicating the voice of the speaker, and records owner information provided near the microphone to identify the speaker. Identification processing for outputting identification information based on the owner information read from the recorded recording medium, identification information addition processing for adding identification information to audio information, and transmission processing for transmitting audio information with identification information added A program for causing a computer to execute the above is provided.

  Furthermore, the present invention collects a speaker's voice and outputs voice information indicating the speaker's voice, a voice recognition process for recognizing the speaker's voice, and a recognition result of the voice recognition process. Provides a program for causing a computer to execute identification processing for determining identification information based on the identification information, identification information addition processing for adding identification information to voice information, and transmission processing for transmitting voice information to which identification information is added. .

  Further, the present invention provides a reception process for receiving voice information to which identification information for identifying a speaker is added, a feature addition process for adding a feature to the voice indicated by the voice information based on the identification information, and a feature addition Provided is a program for causing a computer to execute a sound output process for outputting a sound with a feature added by the process.

  According to the present invention, it is possible to realize an audio conference system in which a listener can easily identify a speaker in an audio conference in which a plurality of people participate from one point. Further, according to the present invention, the speaker can be easily identified even when the audio data recorded in the audio conference is reused.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
[1 First Embodiment]
First, a first embodiment of the present invention will be described. An audio conference system according to the present invention includes an identification unit that outputs identification information for identifying a speaker, an identification information addition unit that adds identification information for identifying a speaker to audio data, and a voice data different for each speaker. The feature adding means for performing the feature adding process for adding the feature is a main component. In the first embodiment, an RFID (Radio Frequency IDentification) tag and an RFID tag reader are used as identification means, and an equalizer is used as characteristic addition means.

[1-1 Configuration of the audio conference system]
FIG. 1 is a block diagram showing a configuration of an audio conference system 100 in the present embodiment. As shown in FIG. 1, the audio conference system 100 according to the present embodiment enables an audio conference between conference rooms A and B located geographically apart from each other.
Each main unit 200 installed in each conference room is connected via the communication network 300 and has a function of transmitting and receiving audio data between the conference rooms. When voice data to be transmitted to another conference room is generated, the main unit 200 identifies the speaker that is the generation source, and adds the identification information of the speaker to the voice data. The main unit 200 performs a feature addition process based on the identification information when the audio data to which the identification information is added is received from the main unit 200 in another conference room. Details of the feature addition processing will be described later.
The communication network 300 is, for example, the Internet. Communication network 300 is not limited to the Internet, and may be another communication network capable of packet communication, such as a wireless communication network or a dedicated line.

  Conference rooms A and B are provided with microphones 10. The microphone 10 is a device that collects voices of conference attendees. In the present embodiment, the microphone 10 has directivity and collects only the voices of attendees sitting in front of the microphone 10. In the present embodiment, microphones 10 are prepared for at least the capacity of the conference room, and attendees can use one microphone per person. Hereinafter, a state where a directional microphone is used and each attendee can use one microphone is referred to as an “on-mic state”. In this on-microphone state, each microphone 10 collects only the sound of one attendee.

  An RFID tag reader 20 is attached to each microphone 10. The RFID tag reader 20 is a device that reads information from an RFID tag owned by an attendee seated in front of the microphone 10. RFID is an individual recognition technology for recognizing an owner in a non-contact manner using radio waves. The RFID system includes an “RFID tag” that is an information recording medium, and an “RFID tag reader” that reads information previously written to the RFID tag by the RFID tag writer. In this system, information in the RFID tag is read by bringing the RFID tag and the RFID reader close to each other within a certain distance range. The present invention can also be implemented using a technology such as a bar code or a magnetic card instead of the RFID system. However, the RFID system can update and add information compared to technologies such as barcodes, can recognize multiple individuals at once, is transparent (can be recognized even in a bag), etc. It has features and is suitable as the identification means of the present invention.

  The microphone 10 and the RFID tag reader 20 are all connected to the main unit 200. Based on the data sent from the RFID tag reader 20, the main unit 200 performs identification processing for identifying the speaker that is the source of the audio signal sent from the microphone 10, and identifies the speaker. It has a function of executing speaker information addition processing for adding information to a voice signal and transmitting the information to a communication partner via the communication network 300.

  When receiving audio data, the main unit 200 corrects the audio data transmitted from the main unit 200 of another conference room via the communication network 300 using the equalizer 270. Thereafter, the corrected audio data is converted into an audio signal, and the audio signal is output to the speaker 30. The speaker 30 is a device that reproduces the audio signal thus given as sound, and is appropriately arranged in each conference room.

  FIG. 2 is a block diagram showing the configuration of the main unit 200 installed in each conference room. The main unit 200 is connected to the communication network 300 shown in FIG. 1 via the communication I / F 240 and can communicate with other communication devices via the communication network 300. The CPU 210 is a processor that controls this communication and performs various arithmetic processes. The memory 250 functions as a work area for the CPU 210, and stores an attendee table TBL1 and a communication partner table TBL2, which will be described later. The audio signal receiving unit 230 has a function of receiving an analog audio signal (hereinafter simply referred to as an audio signal) from the microphone 10 shown in FIG. The CODEC 220 has a function of converting an audio signal output from the microphone 10 into digital audio data (hereinafter simply referred to as audio data) and converting audio data received via the communication network 300 into an audio signal. The equalizer 270 has a function of correcting the sound quality of the sound obtained when the sound data is reproduced by increasing or decreasing a specific frequency component in the sound data received via the communication network 300. The audio data corrected by the equalizer 270 is converted into an audio signal by the CODEC 220 and then output as audio from the speaker 30 shown in FIG. 1 via the audio signal output unit 260.

[1-2 Operation of the audio conference system]
The operation of the audio conference system according to this embodiment will be described below with reference to FIGS. In this embodiment, as shown in FIG. 3, it is assumed that an audio conference is performed between a conference room A at a business office A and a conference room B at a business office B. Participants in the meeting at the A office side are five employees belonging to the C department (employee C1, employee C2, employee C3, employee C4, employee C5), and attendees at the meeting at the B office side are in the D department. Two employees (employee D1, employee D2) who belong and two employees (employee E1, employee E2) who belong to the E section. In this company, each employee uses an employee ID card with an RFID tag on which the employee number of each person is recorded, and all attendees of the meeting have their own ID card in their chest pocket. And

As shown in FIG. 3, the conference room A and the conference room B have a capacity of six people, and therefore, six microphones 10 and six RFID tag readers 20 are prepared. In each conference room, an ID number is assigned to each microphone 10 in advance, and the main unit 200 can distinguish each microphone 10 in each conference room.
Further, the main units 200 of both conference rooms store in advance a sufficient number (for example, six) of correction patterns (hereinafter referred to as equalizing patterns) for increasing or decreasing specific frequency components of the audio data. Each equalizing pattern is assigned an equalizing pattern number for referring to it.

  First, one of the attendees in the conference room A operates the operation panel of the main unit 200 and inputs the conference room number assigned to the conference room B. Then, the main unit 200 in the conference room A transmits a connection request to the main unit 200 in the conference room B via the communication network 300. The main unit 200 in the conference room B that has received the connection request from the main unit 200 in the conference room A opens a communication line with the conference room A. When it is confirmed that both attendees are all seated in front of the microphone 10 (at this time, each attendee is seated in the seat shown in FIG. 3), one of the attendees is Then, the “Attendee Registration” button on the operation panel of the main unit 200 is pressed. By this operation, the operation mode of the audio conference system 100 becomes an attendee registration mode for registering attendees of the conference.

  When the attendee registration mode is set, the main units 200 in both the conference room A and the conference room B command the RFID tag reader 20 attached to the microphone 10 to read the data of the RFID tag. Each of the RFID tag readers 20 reads the employee number from the RFID tag of the attendee sitting in front of the RFID tag reader 20 and transmits the read employee number to the main unit 200. At this time, the RFID tag reader 20 is configured to receive only a signal having an intensity equal to or higher than a predetermined threshold, and is seated in front of the microphone 10 to which the RFID tag reader 20 is attached. Information is not read from RFID tags other than those who participate. The vacant RFID tag reader 20 transmits a signal indicating a vacant seat to the main unit 200. When the main unit 200 receives the employee number from each RFID reader 20, the main unit 200 associates each employee number with the ID number of the microphone 10 to which the RFID reader 20, which is the transmission source, is attached. It is stored in the memory 250 as TBL1.

  In the following description, an aspect using an RFID tag will be described, but the method for identifying a speaker is not limited to the method using RFID. Instead of the non-contact type RFID tag, the identification information of the speaker may be input using a contact type magnetic card or a bar code. In this case, it is necessary to provide the main unit 200 with a magnetic card reader or a barcode reader. Alternatively, the main unit 200 may be provided with an input device such as a keyboard, and an employee number or name may be input by an operation input by an attendee.

  Subsequently, the main units 200 of both conference rooms transmit the attendee table TBL1 of the own room to the main unit 200 of the communication partner. The main unit 200 that has received the attendee table TBL1 of the communication partner assigns a different equalizing pattern number to each employee number described in the attendee table TBL1, and communicates by associating the employee number with the equalizing pattern number. It is stored in the memory 250 as a partner attendee table (hereinafter referred to as “communication partner table TBL2”). This completes preparation for the audio conference. When the preparation is completed, the main unit 200 informs the attendee that the preparation of the meeting is completed by means such as lighting a lamp on the operation panel. The attendee of the meeting confirms this and presses a “start meeting” button on the operation panel of the main unit 200. By this operation, the operation mode of the main unit 200 becomes the conference mode, and the conference can be started.

  Here, consider a case where employee C1 makes a statement in the conference mode. In the present embodiment, since the on-mic state is realized, there is no possibility that the voices of a plurality of speakers are mixed at the time when a speech is collected. The microphone 10 having the ID number 01 that has collected the voice of the employee C1 transmits the voice of the employee C1 to the main unit 200 as a voice signal. The audio signal transmitted from the microphone 10 is converted into audio data by the CODEC 220 in the main unit 200. The CPU 210 of the main unit 200 extracts the employee number (employee number of the employee C1) associated with the microphone 10 with the ID number 01 from the attendee table TBL1, and adds it to the voice data as a header. The voice data to which the employee number is added is transmitted to the main unit 200 in the conference room B, which is a communication partner, via the communication network 300. In the present embodiment, since the Internet is used, a plurality of audio data can be transmitted through one communication line.

  When the CPU 210 of the main unit 200 in the conference room B receives the voice data to which the employee number is added, the CPU 210 extracts the employee number data therefrom. Subsequently, the communication partner table TBL2 in the memory 250 is searched, and the equalizing pattern number corresponding to the employee number is extracted. The audio data is transmitted to the equalizer 270 together with the equalizing pattern number. The equalizer 270 reads the equalizing pattern associated with the equalizing pattern number, and corrects the audio data using the read equalizing pattern.

The main unit 200 outputs the audio data thus processed to the plurality of speakers 30 as audio signals via the CODEC 220 and the audio signal output unit 260, and the audio signals corrected by the equalizer 270 are reproduced from the speakers 30. Thus, by reproducing the sound in which the specific frequency component is changed for each speaker, the listener can easily distinguish the speaker.
For example, assume that the voices of employee C1 and employee C2 are very similar and difficult to distinguish. At this time, even if the discussion heated up and employee C1 said "I agree" and employee C2 said "I disagree" almost simultaneously, each voice is a microphone assigned to each speaker. 10, the employee numbers of the employees C1 and C2 are added to the header and transmitted to the main unit 200 in the conference room B. In the conference room B, the voice of the employee C1 and the voice of the employee C2 are corrected with separate equalizing patterns and reproduced. For this reason, the attendee in the conference room B can easily recognize which of the employee C1 and the employee C2 is in favor and which is against.

  In the embodiment described above, the main unit 200 of each conference room associates the employee number with the microphone ID before starting the audio conference, and the microphone that collects the voice of the speaker during the audio conference. The speaker's employee number was identified based on the microphone ID. However, there is a mode in which the use of such a microphone ID and employee number is unnecessary. First, when an attendee speaks in a conference room, the RFID tag reader 20 reads an RFID from the speaker's RFID tag at the same time. The read RFID is transmitted from the main unit 200 to the main unit 200 in another conference room. The main unit 200 of another conference room that has received the RFID searches the attendee table TBL1 using this RFID as a key, and extracts a corresponding equalizing pattern number. According to this aspect, since information is read from the RFID each time a speech is made, the speaker can be correctly recognized even if participants change their seats during the meeting.

[2 Second Embodiment]
Subsequently, a second embodiment of the present invention will be described. FIG. 4 is a block diagram showing the configuration of the audio conference system according to the second embodiment of the present invention. In the present embodiment, a recording server 320 is connected to the communication network 300 as a device for recording the contents of the conference. In the present embodiment, a URI (Uniform Resource Identifier) that identifies a resource storing various information related to the speaker is used as information for identifying the speaker. In this embodiment, an information server 310 is connected to the communication network 300 as a server that provides this information. Here, the URI is a character string for identifying a resource having a unified format. In the present embodiment, it specifically refers to a URL (Uniform Resource Locator).

[2-1 Configuration of the audio conference system]
In the present embodiment, the configuration of each device provided in each conference room is the same as that of the first embodiment, and thus the description thereof is omitted.
As shown in FIG. 5, the information server 310 records information about attendees in advance, such as a face photo, department, employee number, telephone number, e-mail address, and URI indicating the location where information about the employee is stored. A server having a registered directory database. The recording server 320 is a server for recording audio data transmitted and received during the conference. Each of the information server 310 and the recording server 320 is a server device having an external storage device such as a CPU, a memory, and a hard disk, and a means for performing communication via a network. In FIG. 4, a common information server is provided for the conference room A and the conference room B. However, an information server for each of the conference rooms A and B may be provided separately. Further, the information server 310 and the recording server 320 may be the same device. Further, the main unit 200 itself may have the same function as the information server 310 and / or the recording server 320.

[2-2 Operation of the audio conference system]
The operation of the audio conference system according to this embodiment will be described below with reference to FIGS. Also in the present embodiment, as in the first embodiment, an audio conference is performed between the conference room A of the A office of a company and the conference room B of the B office. Assume that the attendees are the same as those in the first embodiment. When all attendees are gathered in both the conference rooms, the operation mode of the main unit 200 is changed to the participant registration mode when the “attendant registration” button on the operation panel of the main unit 200 is pressed. Hereinafter, the operation up to storing the attendee table TBL1 in the memory 250 is the same as that of the first embodiment.

  After storing the attendee table TBL1 in the memory 250, the main unit 200 transmits the URI request and the attendee table TBL1 to the information server 310. Receiving the URI request, the information server 310 searches the own directory database and extracts the URI corresponding to the employee number described in the attendee table TBL1. The extracted URI is associated with the employee number and added to the attendee table TBL1. The attendee table TBL1 to which the URI is added is transmitted to the main unit 200 that is the transmission source of the URI request. The main unit 200 stores the received attendee table TBL1 in the memory 250 as a new attendee table TBL1. Subsequently, the main units 200 of both conference rooms transmit the attendee table TBL1 to the main unit 200 of the communication partner. The main unit 200 that has received the attendee table TBL1 from the communication partner associates the equalizing pattern number described in the first embodiment with each URI in the received attendee table TBL1, and stores the memory 250 as the communication partner table TBL2. To remember. This completes preparation for the audio conference.

Meeting attendees confirm that they are ready and start the meeting. When the “conference start” button on the operation panel of the main unit 200 is pressed, the main unit 200 transmits a signal indicating the attendance registration completion to the recording server 320 via the communication network 300. When the recording server 320 receives the signal, the recording server 320 creates a file for recording, and automatically assigns a file name that can be distinguished from others using information such as the date and time of the meeting.
This recording file has a file format capable of multitrack recording and reproduction. One track is assigned to each conference participant, and each track is recorded with the voice data of a specific speaker along with a header, and a specific utterance of a specific speaker can be extracted. Note that the format of the recording file is not limited to the multitrack format, and any format may be used as long as it can be stored in a state where speakers and utterances can be identified.

When the attendee speaks, the speech of the meeting attendee is transmitted to the main unit 200 as an audio signal through the microphone 10 assigned to each attendee, as in the first embodiment. The audio signal input from the microphone 10 is converted into audio data by the CODEC 220 in the main unit 200. The main unit 200 extracts the URI associated with the microphone 10 with the ID number 01 from the attendee table TBL1, and adds the extracted URI to the audio data as a header.
As described above, the main unit 200 transmits the audio data with the URI added to the main unit 200 and the recording server 320 in the conference room B via the communication network 300. The audio reproduction in the conference room B is the same as that in the first embodiment except that the “employee number” in the first embodiment is changed to “URI”.
When receiving the audio data, the recording server 320 records the received audio data in the recording file. The URI added as a header is also recorded at the same time. Thus, the contents of the conference are recorded in the recording file of the recording server 320.

Here, it is assumed that after a meeting, one of the attendees of the meeting listens again to a file in which the contents of the meeting are recorded, for example, in order to create a minutes. The employee who creates the minutes first accesses the recording server 320 from his / her terminal (not shown) via the communication network 300, and opens the file in which the previous meeting was recorded. The terminal is preinstalled with playback software for playing back files recorded by the voice conference system.
When playing back the recorded file, this playback software detects the audio data in each track, and accesses the resource indicated by the URI added as a header (a file describing information about the employee C1 in the information server 310). Get data for the name. The reproduction software reproduces the voice data while characterizing each speaker by displaying the name of the speaker obtained from the information server 310 on the screen. Note that the method for characterizing each speaker at the time of reproducing the sound data is not limited to the display of the name, and the sound waveform may be displayed in a different color for each speaker. Or it is good also as a structure which reproduces | regenerates audio | voice data, displaying a speaker's face photograph instead of a speaker's name. Alternatively, information such as the speaker's employee number and e-mail address may be displayed. Alternatively, an equalizer that increases / decreases frequency components that are different for each speaker may be applied when reproducing a recording, as in the case of an audio conference. Alternatively, sound may be reproduced from a speaker / headphone with different sound localization for each speaker.
According to the present embodiment, the user who is listening to the recording file of the audio conference can hear the recording of the conference without wondering who the speaker is or without misidentifying the speaker.

[3 Third Embodiment]
Subsequently, a third embodiment of the present invention will be described. In this embodiment, a speech recognition technique is used as the identification means, and a technique for changing the localization of the sound image for each speaker using a plurality of speakers as the feature addition means. In the following description, the same reference numerals are used for the same components as those in the first embodiment and the second embodiment.

[3-1 Configuration of audio conference system]
FIG. 6 is a block diagram showing a configuration of an audio conference system according to the third embodiment of the present invention. In the present embodiment, a main unit 500 is used instead of the main unit 200 used in the first embodiment and the second embodiment.
FIG. 7 is a block diagram showing a configuration of the main unit 500 in the present embodiment. In FIG. 7, a voice recognition unit 550 and a sound image localization processing unit 570 are components unique to the present embodiment. The voice recognition unit 550 has a function of performing voice recognition processing on the voice signal input from the microphone 10 shown in FIG. The sound image localization processing unit 570 has a function of performing processing for changing the sound image localization of the sound reproduced from the speaker 30. Here, the position where the sound image is localized is determined by the sound image localization information stored in the sound image localization processing unit 570. The sound image localization processing unit 570 stores in advance a sufficient number (six in this embodiment) of sound image localization information (for example, distance and angle from a certain reference point), each of which refers to the sound image localization information. A sound image localization number is assigned. Assume that the sound image is localized at the positions shown in FIG. In this case, the sound image localization processing unit 570 has six pieces of sound image localization information including combinations of distances and angles from the reference point, and sound image localization numbers 1 to 6 are assigned to each. That is, when the sound data and the sound image localization number are given, the sound image localization processing unit 570, based on the sound image localization information corresponding to the sound image localization number, the phase of the sound output from each of the speakers 30 shown in FIG. The sound image is localized by changing the intensity.

[3-2 Operation of the audio conference system]
The operation of the audio conference system according to this embodiment will be described below with reference to FIGS. In starting the conference, one of the attendees presses the “Register Attendee” button on the operation panel of the main unit 500. By this operation, the operation mode of the main unit 500 becomes the attendee registration mode. In the present embodiment, each attendee gives his / her name to the microphone 10, and the main unit 500 performs speech recognition processing on the speech and obtains the attendee's name to perform speaker identification. Details are as follows.

When the employee C1 gives his name “Ichiro Suzuki” to the microphone 10, the sound is transmitted to the main unit 500 through the microphone 10 as a sound signal. When the main unit 500 receives the audio signal, the main unit 500 transmits the received audio signal to the audio recognition unit 550. The voice recognition unit 550 performs voice recognition processing on the voice signal, and extracts text data named “SUZUKI Ichiro”. The main unit 500 transmits the extracted text data and URI request to the information server 310 via the communication network 300.
Upon receiving the name text data and the URI request, the information server 310 searches the name list database from the name list database (FIG. 5) held by the information server using the name text data as a key. When the information server 310 detects employee data corresponding to the text data “Suzuki Ichiro”, the information server 310 extracts a URI indicating the location where the data related to the employee is stored from the name list database, and returns the URI to the main unit 500. If the same name is used, information other than the name such as “C Section Ichiro Ichiro” may be added when the name is given, and the database may be searched by name and department. Alternatively, if there is an employee with the same name and the same name, the information server 310 transmits a list of the names and employee numbers of the employees with the same name to the main unit 200, and the employee C1 is displayed from the list displayed on the display of the main unit 200. You may be allowed to select your own data.
When the main unit 500 receives the URI, the main unit 500 stores the received URI and the ID number of the microphone 10 of the employee C1 in association with each other as the attendee table TBL1.

  By executing the above processing for all attendees, the main unit 500 of each conference room, for all attendees of the conference, the ID number of the microphone 10 at the seat where each attendee sits and the attendance thereof. The attendee table TBL1 associated with the URI indicating the place where the information on the person is stored is stored. When the creation of the attendee table TBL1 relating to the attendees in the conference room in which the conference room is installed is completed, the main unit 500 in both conference rooms transmits the attendee table TBL1 held by the communication partner. The main unit 500 that has received the communication partner attendee table TBL1 associates the sound image localization number with each URI recorded in the received attendee table TBL1, and stores it as the communication partner table TBL2. This completes the attendee registration mode. After the attendee registration mode ends, the mode shifts to the normal conference mode. In the normal conference mode, the audio conference system operates as follows.

  When the employee C1 speaks, the comment is transmitted to the main unit 500 through the microphone 10 as an audio signal. The received audio signal is converted into audio data by the CODEC 220 in the main unit 500, and the URI (in this case, employee C1) associated with the ID number of the microphone that collected the audio signal in the converted audio data. (URI indicating the storage location of the information) is added as a header. The audio data to which the URI is added is transmitted to the communication destination main unit 500 via the communication network 300. The main unit 500 that has received the audio data to which the URI is added searches the communication partner table TBL2 stored in its own memory to see if the URI is registered, and the sound image stored in association with the URI. Extract localization information.

The main unit 500 adds the sound image localization information to the header of the audio data, and transmits the audio data to the sound image localization processing unit 570 in the main unit 500. The sound image localization processing unit 570 determines the sound image localization of the sound based on the sound image localization information added to the header, and corrects the audio signal so that the sound is reproduced with the sound image localization. The corrected audio signal is reproduced as audio from the speaker 30. According to the present embodiment, since the sound is reproduced from the plurality of speakers 30 with different sound image localization for each speaker, the attendee of the conference can easily identify the speaker.
In the present embodiment, sound is reproduced from a plurality of speakers 30 with different sound image localization for each speaker. However, speakers 30 are prepared for the number of participants in the conference, and one speaker is assigned to each speaker. The voice of each speaker may be reproduced from a speaker specified for each speaker. Further, a display may be provided on the main unit 200 so that the employee number or name of the speaker who is speaking is displayed on the display.

[4 Fourth Embodiment]
Subsequently, a fourth embodiment of the present invention will be described. FIG. 8 is a diagram showing a configuration of an audio conference system according to the fourth embodiment of the present invention. The voice conference system according to this embodiment is different from the first to third embodiments described above in that the number of microphones 10 in the conference room is smaller than the number of conference rooms (hereinafter referred to as “off-microphone state”). . In the present embodiment, the microphone 10 is an omnidirectional microphone, and the speech of a plurality of speakers is collected from the plurality of microphones 10 as mixed sound. The sound collected from the microphone 10 is first separated for each sound source, and then speaker recognition processing is performed on the separated sound. In the following description, the same components as those in the first to third embodiments are denoted by the same reference numerals.

[4-1 Configuration of the audio conference system]
As shown in FIG. 8, the audio conference system 700 in this embodiment includes a communication network 300, a main unit 800, an information server 310, a microphone 10, and a speaker 30. In the present embodiment, the name database extracted from the voice signal indicating the voice of each employee with his / her name is recorded in the name list database stored in the information server 310 in advance.

  FIG. 9 is a block diagram showing a configuration of the main unit 800. The sound image localization measuring unit 910 has a function of measuring the sound image localization of the sound source of the sound input from the three microphones 10 shown in FIG. The main unit 800 stores sound image localization information of the sound source measured by the sound image localization measuring unit 910. The speaker recognizing unit 950 performs processing for recognizing the speaker for the audio data for each speaker separated by the sound source separation unit 900 described later, and adds information for identifying the speaker to the audio data. Have The sound source separation unit 900 has a function of performing sound source separation processing on sound signals input via the plurality of microphones 10 based on sound image localization information of the sound source, and separating the sound signals for each speaker. Specifically, the sound source separation unit 900 includes a synchronous addition unit composed of three delay units that change the phases of the audio signals input from the three microphones 10 shown in FIG. It has six. One sound source is assigned to each synchronous adder. Each synchronous adder has a function of correcting the phase difference estimated from the sound image localization information (angle) of the corresponding sound source, and adding the sound signals from the sound source in phase. As a result, the sound signal from the specific sound source is in phase, but the signals from the other sound sources are not in phase, so only the sound signal from the specific sound source is emphasized, and the sound source can be separated. Note that the sound source separation method is not limited to this method, and a blind sound source separation (BSS) method based on independent component analysis (ICA), a separation method assuming a harmonic structure of speech, or the like. Other techniques may be used.

[4-2 Operation of the audio conference system]
Hereinafter, the operation of the audio conference system according to the present embodiment will be described with reference to FIGS. 8 and 9. First, attendee registration is performed in the attendee registration mode. Attendees will give their names in turn. The voice uttered by the attendee is transmitted to the main unit 800 as a voice signal through the plurality of microphones 10. The sound image localization measuring unit 910 calculates the angle of the sound source (speaker) with respect to the microphone 10 from the phase differences of the plurality of audio signals output from the plurality of microphones 10. An ID number is assigned in advance to each synchronous adder of the sound source separator 900.

  The speaker recognizing unit 950 extracts voice features by processing such as analyzing the spectrum and formants of the voice data and extracting feature amounts. Subsequently, the speaker recognition unit 950 transmits the speaker identification request and the extracted feature amount to the information server 310 via the communication network 300. When the information server 310 receives the speaker identification request, the information server 310 searches the name list database held by the information server 310 based on the received feature quantity, and identifies the speaker as the employee C1 from the feature quantity. When the speaker is identified, the information server 310 transmits information for identifying the speaker to the main unit 800, such as the employee number of the employee C1 or a URI indicating the location where the information related to the employee C1 is stored. The main unit 800 that has received the information for identifying the speaker associates the received URI, the sound image localization information (angle) of the speaker previously calculated, and the ID number of the vacant synchronization adding unit with each other in association with the attendee table. Store as TBL1. Each synchronous adder performs a synchronous addition process based on sound image localization information associated with its own ID number. The operation in which the main units 800 in both conference rooms transmit the attendee table TBL1 to the communication partner main unit 800 to create the communication partner table TBL2 is the same as in any of the first to third embodiments. This completes the attendee registration mode.

  In the conference mode, when an attendee (for example, employee C1) speaks, the voice of the employee C1 is transmitted to the main unit 800 as a plurality of audio signals via the plurality of microphones 10. The main unit 800 transmits each received audio signal to the sound source separation unit 900 in the main unit 800. As described above, the sound source separation unit 900 can extract voice data for each speaker. The sound source separation unit 900 extracts, from the attendee table TBL1, the URI associated with the ID number of the synchronous addition unit to which the audio data is output from the extracted audio data. The extracted URI is added to the audio data as a header. The voice data to which the information for identifying the speaker is added as described above is transmitted to the communication partner main unit 800 via the communication network 300. Hereinafter, the reproduction processing in the communication destination conference room and the reuse of the audio data after the conference are the same as the reproduction processing and the audio data reuse in the first to third embodiments described above.

[5 Modification]
The above-described embodiment is an exemplification of the embodiment of the present invention, and various modifications can be made to the above-described embodiment without departing from the gist of the present invention.

[5-1 First Modification]
As described above, the present invention is generally composed of the following three elements.
(1) Identification means (2) Means for adding speaker identification information to voice data (3) Feature addition means Specific combinations of the elements (1) to (3) in the first to fourth embodiments Although an aspect and the modification were illustrated, the combination is not limited to the combination described in each embodiment, Arbitrary combinations including a modification are possible.

[5-2 Second Modification]
In each of the above-described embodiments, the audio conference between two points has been described. However, a configuration in which an audio conference is performed between three or more points may be employed. In this case, in the attendee registration mode, the main unit transmits the attendee table TBL1 to a plurality of communication partners and stores the communication partner table TBL2 of the plurality of communication partners.

[5-3 Third Modification]
In the fourth embodiment, speaker recognition is performed after sound source separation in the off-mic state, but an RFID tag is used supplementarily when performing speaker recognition, and a database that is searched for speaker recognition is used. It is good also as a structure which narrows down. In this case, the operation of the audio conference system is as follows.

  An RFID tag reader 20 is attached to the microphone 10. When the attendee of the conference presses the “register attendee” button on the operation panel of the main unit 800, the main unit 800 reads the employee number of the attendee of the conference from the RFID tag reader 20. Command. Since the RFID tag can be read at a certain distance, the RFID tag reader 20 attached to the microphone 10 in the conference room A is used by employees (employee C1, employee C2, employee C3) present in the conference room A. The employee number data is read out from the RFID attached to the employee ID of employee C4 and employee C5). The read employee number data is transmitted to the main unit 800. The main unit 800 transmits the voice feature amount request together with the employee number data to the information server 310. The information server 310 that has received the voice feature value request searches its own speaker information database using the received employee number as a key, and obtains the voice feature values of the employee C1, employee C2, employee C3, employee C4, and employee C5. Extract. The information server 310 transmits the extracted audio feature quantity to the main unit 800. When the main unit 800 receives the voice feature amount, the main unit 800 associates and stores the employee number and the voice feature amount of each attendee as a voice recognition table. The meeting is now ready.

  During the conference, the speaker recognition unit 950 that has received the voice data extracts voice features by processing such as analyzing the spectrum and formants and extracting feature amounts from the voice data as described in the fourth embodiment. . Subsequently, the speaker recognition unit 950 searches the speech recognition table using the extracted feature amount as a search key, and extracts the employee number of the speaker. The processing subsequent to adding the extracted employee number to the voice data is the same as that in the fourth embodiment.

[5-4 Fourth Modification]
In the fourth embodiment, voice recognition of each attendee is performed in the attendee registration mode, and the sound image localization information obtained as a result and the URI of each attendee are associated and recorded in the attendee table TBL1. However, in the attendee registration mode, each attendee is not associated with sound image localization information, and in the conference mode, speaker recognition processing is performed for each utterance, and speaker identification information (URI or employee number) is added to the voice data. It is good also as a structure.

It is a block diagram which shows the structure of the audio conference system in 1st Embodiment of this invention. It is a block diagram which shows the structure of the main unit in the embodiment. It is a schematic diagram which shows the structure of the audio conference system in the embodiment. It is a block diagram which shows the structure of the audio conference system in 2nd Embodiment of this invention. It is a figure which illustrates the name list database recorded on the information server in the embodiment. It is a block diagram which shows the structure of the audio conference system in 3rd Embodiment of this invention. It is a block diagram which shows the structure of the main unit in the embodiment. It is a figure which shows the structure of the audio conference system in 4th Embodiment of this invention. It is a block diagram which shows the structure of the main unit in the embodiment. It is a figure which illustrates the sound image localization position in 3rd Embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 ... Microphone, 20 ... RFID tag reader, 30 ... Speaker, 100 ... Voice conference system, 200 ... Main unit, 210 ... CPU, 220 ... CODEC, 230 ... Voice signal receiving part, 240 ... Communication I / F, 250 ... Memory, 260 ... Audio signal output unit, 270 ... Equalizer, 300 ... Communication network, 310 ... Information server, 320 ... Recording server, 400 ... Terminal, 500 ... Main unit, 550 ... Voice recognition unit, 570 ... Sound image localization processing unit, 700 ... Voice conference system, 800 ... Main unit, 900 ... Sound source separation unit, 910 ... Sound image localization measurement unit, 950 ... Speaker recognition unit

Claims (6)

A microphone that collects the voice of the speaker and outputs voice information indicating the voice of the speaker;
An identification unit that is provided in the vicinity of the microphone and outputs identification information based on the owner information read from a recording medium in which owner information for identifying the speaker is recorded;
Identification information adding means for adding the identification information to the voice information;
A voice conference terminal device comprising: transmitting means for transmitting the voice information to which the identification information is added. A microphone that collects the voice of the speaker and outputs voice information indicating the voice of the speaker;
Voice recognition means for recognizing the voice of the speaker;
Identification means for determining identification information based on a recognition result of the voice recognition means;
Identification information adding means for adding the identification information to the voice information;
A voice conference terminal device comprising: transmitting means for transmitting the voice information to which the identification information is added. Receiving means for receiving voice information to which identification information for identifying a speaker is added;
A feature adding means for adding a feature to the voice indicated by the voice information based on the identification information;
A voice conference terminal device comprising: voice output means for outputting the voice with the feature added by the feature adding means. Voice information output processing for collecting the voice of the speaker and outputting voice information indicating the voice of the speaker;
An identification process that is provided in the vicinity of the microphone and outputs identification information based on the owner information read from the recording medium in which the owner information for identifying the speaker is recorded;
An identification information adding process for adding the identification information to the voice information;
A program for causing a computer to execute transmission processing for transmitting the audio information to which the identification information is added. Voice output processing for collecting the voice of the speaker and outputting voice information indicating the voice of the speaker;
A voice recognition process for recognizing the voice of the speaker;
An identification process for determining identification information based on a recognition result of the voice recognition process;
An identification information adding process for adding the identification information to the voice information;
A program for causing a computer to execute transmission processing for transmitting the audio information to which the identification information is added. A receiving process for receiving voice information to which identification information for identifying a speaker is added;
A feature addition process for adding a feature to the voice indicated by the voice information based on the identification information;
A program for causing a computer to execute a sound output process for outputting a sound with a feature added by the feature addition process.

Images