JP2009027239A - Telecommunication conference apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a telecommunication conference apparatus which can make it easy to hear the speaker's voice by judging the speaker of a telecommunication conference system more naturally. <P>SOLUTION: The telecommunication conference apparatus in a telecommunication conference system connected with a plurality of teleconferencing terminals provide a comfortable teleconferencing environment by monitoring the volume of meeting participants in order to specify a speaker, and then increasing the volume of the speaker while suppressing unnecessary voice of non-speakers. The break of utterance of the speaker is judged by monitoring the silent time duration and unnatural volume control by superfluous control is suppressed. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a communication conference device that is used, for example, when a communication conference by voice call is performed by three or more people and used for a conference call, a remote video conference, and the like.

  Patent Document 1 discloses a telephone conference system. When the telephone conference system talks with a plurality of people using a handset of a single telephone device, it feels different from normal conversation and communicates smoothly. The problem that cannot be achieved is solved. In addition, as the number of people who talk at the same time increases, it becomes difficult to distinguish the story of the other party who is the center of the conversation and solves the problem of poor usability. Specifically, the following measures are taken.

-It is determined as a speech to be discussed by the ratio of the voice overlap time T1 and the speech time T2 of the voice,
Control the gain control amplifier with weight W = T1 / T2.
JP 2000-49948 A

However, the telephone conference system described in Patent Document 1 has the following problems.
・ When T1 and T2 are measured and W is calculated, the discussion is already in progress.
-In orderly discussions, remarks do not overlap, and T1 is not necessarily the judgment index for the speech of interest.
In other words, there may be a time lag between the noticed voice and the actual speaker, which is not necessarily a natural conversation.

  An object of the present invention is to provide a communication conference device that makes it easier to hear a speaker's voice by making the determination of the speaker more natural.

  The communication conference apparatus of the present invention is a communication conference system in which a plurality of audio conference terminals are connected, and among the plurality of audio conference terminals, the function of specifying an audio conference terminal in which a participant of the communication conference is speaking, A function for varying the call volume for each voice conference terminal in which the participant is speaking, a function for determining that the participant has finished speaking, and a function for emphasizing the voice conference terminal in which the participant is speaking It is characterized by having.

  According to the present invention, it is possible to realize a more natural communication conference by emphasizing the voice of the speaker who is speaking from the microphone of the audio conference terminal.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

The communication conference apparatus of the present invention determines a speaker by the following configuration.
(1) When a volume higher than the call determination volume is detected when A speaks, it is determined that the communication conference terminal 1a speaks, and the amplifier gains other than the participant A are reduced.
(2) The silence period of A is monitored, and when the time equal to or longer than the silence monitoring time T0 has elapsed, it is determined that the speech has ended, and the amplifier gains other than A are restored. Then wait for the next speaker. The call determination volume and the silent monitoring time T0 will be described later.
(3) Next, when B speaks, it becomes speaker B, and amplifier gains other than speaker B are reduced.

  FIG. 1 is an overall view of a communication conference system using the communication conference apparatus 10 of the present invention. In FIG. 1, 1a, 1b, and 1c are voice conference (general purpose) terminals, 2 is a network / phone line, and 10 is a communication conference device. The conference participants A, B, and C conduct a communication conference via the network / telephone line 2 using the audio conference (general purpose) terminals 1a, 1b, and 1c.

  The network / telephone line 2 is connected to the communication conference device 10 of the present invention, and the communication conference device 10 smoothly promotes a voice communication conference via the network / telephone line 2.

  FIG. 2 is a block configuration diagram of the communication conference apparatus 10 according to the first embodiment of this invention. In FIG. 2, 100 is a network interface input circuit and decomposition circuit, 101 is an input volume adjustment unit, 102 is a volume monitoring unit, 103 is a speech analysis unit, 104 is a gain adjustment unit, 105 is an adder circuit and network interface output circuit, 106 Indicates a central control circuit, and 107 indicates a database.

  Each functional block shown in the block configuration diagram may be realized by a digital signal processing processor or hardware.

  Each functional block will be described below. The network interface input circuit / decomposition circuit 100 receives data including audio data of a plurality of participants in the communication conference from the network, decomposes the data for each audio conference terminal, and outputs the audio to each input audio adjustment unit 101. Pass data.

  Based on the volume data of each audio conference terminal from the volume monitoring unit 102, the input volume adjustment unit 101 adjusts the volume of the input audio from each conference participant to a predetermined predetermined audio level.

  The sound volume monitoring unit 102 analyzes the sound data that has passed the input sound volume adjusting unit 101 and has a constant sound level, and monitors the sound volume.

  The speech analysis unit 103 monitors the voice data that has passed through the input volume adjustment unit 101 and analyzes the speech status of the conference participants. For example, the number of utterances, the time until the next utterance becomes greater than or equal to the call determination volume from the time when the call determination volume or less is reached, the frequency distribution of this time, and the like are analyzed.

  The gain adjustment unit 104 adjusts the amplifier gain of the participant to be emphasized to “large” in accordance with an instruction from the central control circuit in order to distinguish between the participant who emphasizes the speech and the participant who suppresses the speech. By adjusting the amplifier gain to “small”, the speaker's transmission gain is improved.

  The adder and network interface output circuit 105 mixes the conference participant audio data that has passed through the input audio adjustment unit 101 and the gain adjustment unit 104, and sends them to all conference participants by multicast or broadcast.

  The central control circuit 106 determines the presence / absence of the speaker based on information collected from the volume monitoring unit 102 and the speech analysis unit 103, and determines which speech is emphasized when there is a speech, and gain Instructs the adjustment unit 104.

  The database 107 stores the analysis result of the maximum volume, the average volume, the input volume adjustment value, and the occurrence frequency distribution (such as the interval between the statements) for each speaker.

  The basic operation of the input voice and amplifier gain control according to the first embodiment of the present invention will be described with reference to FIG.

In FIG.
(1) State without speech. The amplifier gain of all ABC participants is set to “large”. By setting the amplifier adjustment from “large” to “small”, a participant who is speaking can have a natural conversation.
(2) Since the voice of Mr. A exceeds the “call determination volume Vsh”, the gain of Mr. B and Mr. C is set to “small” to emphasize Mr. A's remarks.

(3) The voice of Mr. A falls below “call determination volume Vsh”, and the silent monitoring period is reached.
(4) Since the voice of Mr. A did not exceed the call determination volume Vsh again during the “silent monitoring time T0”, it is determined that there is no speaker and the gain of all ABC users is “high”. And Here, the silent monitoring time T0 means the silent time of the speaker's breath and the time between the conversations of the speaker.
(5) Since the voice of Mr. B exceeds the call determination volume Vsh, the gain of Mr. A and Mr. C is set to “small” to emphasize Mr. A's remarks.

(6) While Mr. B's speech, Mr. C's voice exceeded the call determination volume Vsh, but Mr. B's voice is above the call determination volume Vsh, so Mr. C's gain remains “small”. do not do.
(7) Mr. B's voice falls below the call determination volume Vsh, and the silent monitoring period is reached.
(8) Since the voice of Mr. B exceeded the call determination volume Vsh again during the silent monitoring time T0, the previous state is continued.

(9) Mr. B's voice falls below the call determination volume Vsh, and the silent monitoring period is entered.
(10) Since the voice of Mr. B did not exceed the call determination volume Vsh again during the silent monitoring time T0, the state of no speaker is set, and the gains of all ABC users are set to “high”.

  FIG. 4 is a flowchart for explaining the operation of generating a database used for the volume adjustment operation for each speaker according to the first embodiment of the present invention.

  When the flowchart of FIG. 4 is started, the process waits for the input voice level (frequency of a human voice) to exceed a predetermined call determination volume (Vsh) (S100). When the input voice level exceeds the call determination volume (Vsh) (S100: Y), the measurement of the input voice is started (S101). Measurement of the input voice continues until the input voice level falls below the call determination volume (Vsh) (S102).

  When the input voice level falls below the call determination volume (Vsh) (S102: Y), the measurement of the input voice level is ended (S103). Then, the measurement data is analyzed, the maximum value and the average value of the input voice level are calculated, and the calculated values are added to the database 107 (S104). An appropriate input volume adjustment value is calculated and updated for each speaker from past data registered in the database 7 and newly added data (105).

  According to the flow in FIG. 4, the volume of each participant may be different at the stage of reaching the conference device due to differences in the environment, equipment, and individual speaker volume of the conference participants. When trying to determine the speaker as it is, for example, a person who speaks louder or a person closer to the microphone wins, and the opportunity to speak cannot be given fairly.

  Therefore, by calculating the input volume adjustment value, the volume that has reached the conference device is appropriately monitored for each participant, and the input volume of the conference device is set so that the volume is equivalent among the conference participants appropriate for speaker determination. adjust. Further, since appropriate volume adjustment is performed at the input stage, the switching of the amplifier gain for emphasizing the speaker can be as simple as switching between two stages of “large” and “small”.

  FIG. 5 is a diagram showing volume adjustment information for each speaker stored in the database 7. The database 107 stores data of participants, maximum volume of input voice level, average volume, call determination volume, and input volume adjustment value.

  FIG. 6 is a flowchart for explaining the operation for adapting the silent monitoring time T0.

  Some people speak at a time in a teleconference, while others speak carefully while thinking one word at a time, and there are individual differences depending on the participants. In the latter case, it is often the case that the silent time exceeds a certain time, but it is erroneously determined that there is no speaker even though the speech is being made.

  By optimizing the silent monitoring time T0, the number of erroneous determinations can be reduced for each participant.

  A speech pattern (a speech-to-speech interval or the like) is appropriately monitored, and a time for determining a break (silent monitoring time T0) is adapted.

  As shown in the flowchart of FIG. 6, first, when participating in a communication conference, a search is made for past participant records in the database for silent monitoring adaptation for each participant. Specifically, although not shown in FIG. 7, information for voiceprint authentication is accumulated for each participant, and the participant is specified by the input voice. (S200)

  When there is past data in the database (S200: Y), the silent monitoring time T0 is set to a value recorded in the database (S201).

  If there is no past data in the database (S200: N), the silent monitoring time T0 is set to the system default value X (for example, 300 mS) (S202). Further, it is added as a new participant to the database (S203).

  Next, it is determined whether or not silence has been detected for each participant (that is, is it below the call determination volume Vsh?) (S204). If no silence is detected, the process is repeated until it is detected.

  When silence is detected (S204: Y), a silence time measurement timer T1 that times out at a predetermined time (for example, 3 seconds) is started in order to determine that the speech has been completed or to measure the silence monitoring time. Then, the silent time measurement is started (S205). Here, when the communication conference is ended (S206: Y), the silent time measurement timer T1 is stopped (S207), and the process is ended.

  Usually, when the communication conference has not ended (S206: N), it is determined whether the silent time measurement timer T1 times out (S208).

  If not timed out (S208: N), it waits for a voice to be input (S209). Here, when the voice of the participant is detected (step S209: Y), the silent time measurement timer T1 is stopped, the value of the time t1 at the time of the stop is analyzed, and the silent monitoring time database 7 shown in FIG. 7 is updated. . Specifically, the frequency of the occurrence frequency distribution corresponding to the timer value is added by one, and the number of samples is incremented by one. Further, the contents of the database are analyzed, the silent time mode value and the silent monitoring time t0 are updated to optimum values, and then the silent time measurement timer T1 is reset, and then the process returns to step S204 to detect silence. wait.

  If the silent time measurement timer T1 has timed out and timed out (S208: Y), the amplifier gain of the participant is adjusted to “large” (step S211), the silent time measurement timer T1 is reset, and then the step S206. Migrate to

  FIG. 7 shows the database 107 used for adapting the silent monitoring time T0 according to the first embodiment of the present invention. In the database 107, the participant, the number of samples, the silent time mode value, the occurrence frequency distribution, and the silent monitoring time T0 are recorded.

FIG. 8 is a diagram for explaining the basic principle of a method for eliminating an unnecessary gain change at the beginning of a statement according to the second embodiment of the present invention.
The first embodiment of the present invention is characterized in that when there is no speaker, the gain of all conference participants is set to “large”, and the volume of the speaker is not changed when the gain is varied for emphasizing the speaker. We are going to lose.

  However, if at least one of the conference participants is participating from an environment with high background noise, when there is no speaker, the gain of all conference participants or participants with high background noise is reduced to “low”. It is better to control, and in that case, the gain at the start of speech changes from “small” to “large”, so that the participant's speech becomes difficult to hear.

  FIG. 9 is a block diagram of the communication conference apparatus according to the second embodiment of the present invention. FIG. 8 shows a configuration in which an audio data buffer 108 is added before each gain adjustment unit 104.

  The voice data buffer 108 temporarily stores the voice in the memory and reproduces it by delaying it for a short time after raising the gain to “high”, thereby eliminating the volume change at the beginning of the voice. Note that the slight delay generated here can be made closer to the actual voice as the utterance progresses, for example, by omitting the reproduction of silence time, and the effect is reduced even in environments where delay is not desirable. it can.

  Referring to the temporal changes in “actual speech” and “reproduced speech” shown in FIG. 9, (1) speech is detected for “actual speech”, and (2) gain for “reproduced speech”. After the gain of the adjustment circuit is changed from “small” to “large”, (3) output of audio data is started. Also, (4) silent sections of “reproduced speech” are thinned out little by little so as to approach “actual speech”. In (5), the start times of “actual audio” and “reproduced audio” are substantially the same.

  If it is reproduced as it is, the sound is output before the gain adjustment of the gain adjustment circuit and becomes unnatural. FIG. 9 shows a position on the block diagram where a buffer is added before each gain adjusting unit 104 (108). Whether or not the reproduced sound approaches the actual sound is determined by the amount of data existing in the buffer. The smaller the amount of data, the closer the timing between the actual voice and the reproduced voice is.

  When adjusting the gains of a plurality of voice conference terminals, speech mode: mode in which the amplifier gain is unlikely to change, discussion mode: mode in which the amplifier gain is likely to change, or remarks by two or more participants It is possible to set the mode to be emphasized, etc. If the speaker keeps speaking for a long time, setting the speech mode or resetting the gain balance can eliminate the disadvantage that other people's voice remains at a low level. it can.

  In addition, it is possible to reset the adjusted gain balance of all the conference participants and to set the gain adjustment of a plurality of audio conference terminals with certainty or every silent period.

  In the present embodiment, the call determination volume Vsh has been described. However, the participant's speech start determination and the participant's speech end determination volume may be separately performed. Specifically, the call start determination volume Vsh1 and the call end determination volume Vsh2 are provided so that the determination volume of the call start determination volume Vsh1 is set low, and the determination volume of the call end determination volume Vsh2 is set high, thereby achieving higher accuracy. The silent monitoring time T0 can be measured.

FIG. 1 is an overall view of a communication conference system using the communication conference apparatus of the present invention. FIG. 2 is a block diagram of the communication conference apparatus according to the first embodiment of the present invention. FIG. 3 is a diagram for explaining the basic principle of the relationship between the input voice and the amplifier gain control according to the first embodiment of the present invention. FIG. 4 is a flowchart for explaining the database creation operation according to the first embodiment of the present invention. FIG. 5 is a diagram illustrating a database for adjusting the volume by appropriately monitoring the volume for each speaker according to the first embodiment of the present invention. FIG. 6 is a flowchart for explaining the operation for adapting the silent monitoring time T0 according to the first embodiment of the present invention. FIG. 7 is a diagram showing a database used for adapting the silent monitoring time T0 according to the first embodiment of the present invention. FIG. 8 is a block diagram of the communication conference apparatus according to the second embodiment of the present invention. FIG. 9 is a diagram for explaining the basic principle of a method for eliminating an unnecessary gain change at the beginning of a statement according to the second embodiment of the present invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1a Voice conference terminal 1b Voice conference terminal 1c Voice conference terminal 2 Network / telephone line 10 Communication conference apparatus 100 Network interface input circuit and decomposition circuit 101 Input volume adjustment part 102 Volume monitoring part 103 Statement analysis part 104 Gain adjustment part 105 Addition circuit and Network interface output circuit 106 Central control circuit 107 Database 108 Buffer

Claims (4)

  A communication conference apparatus in a communication conference system to which a plurality of voice conference terminals are connected, and a function of identifying a voice conference terminal that inputs speech voices of nearby communication conference participants among the plurality of voice conference terminals A function of varying the transmission gain for each voice conference terminal that is inputting speech voices of the nearby communication conference participants, a function of determining that the neighboring communication conference participants have finished speaking, A communication conference apparatus having a function of displaying the voice conference terminal. The teleconference device according to claim 1,
The communication is characterized in that the transmission gain of the audio conference terminal that is inputting the voice of the nearby communication conference participant is set to be larger than that of the audio conference terminal that is not inputting the audio of the nearby communication conference participant and transmitted. Conference equipment. In the communication conference apparatus according to claim 1 or 2,
A communication conferencing apparatus characterized by monitoring a speech voice pattern for each voice conference terminal and determining a break or end of a speech for each voice conference terminal. In the communication conference apparatus according to claim 1,
When the system setting for switching the transmission gain from small to large when speech voice is detected is set, the function for temporarily storing the speech voice and the temporary storage after switching the transmission gain from small to large A communication conference apparatus, comprising: a function of reproducing the speech sound that has been played; and a function of reproducing the silent part every time a silent part is detected after the start of reproduction of the speech sound.

Images