JPH06292198A - Talker deciding system - Google Patents

Abstract

PURPOSE:To obtain a talker deciding system by which no unnatural video switching is performed by selecting a preceding talker preferentially when either of high-order two video conference terminals with high voice detecting volume is the video conference terminal of the preceding talker at the time of video siwtching by talker decision is performed. CONSTITUTION:Preceding talker storing memory 11 which stores the preceding talker, and a high-order two voice count decision part 10 which obtains the high-order two video conference terminals with the high voice detecting volume and decides the video conference terminal of the preceding talker or the video conference terminal with the largest voice detecting volume as a talker based on the difference between those voice detecting volumes are added to a talker decision part 1. Thereby, the talker deciding system which performs no video switching can be obtained when the video conference terminal at a second place out of the high-order two video conference terminals with high voice detecting volume is the talker at the time of switching to a preceding video.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speaker determination system for determining a speaker from voices from a plurality of video conference terminals and distributing the video to other terminals in a multipoint communication control device or the like. Is.

[0002]

2. Description of the Related Art FIG. 18 shows, for example, JP-A-1-28668.
FIG. 9 is an overall structural diagram showing a conventional multipoint communication control device as shown in Japanese Patent Publication No. In the figure, 110 is a multipoint control device, 111 is a video camera, and 112 is a CR.
T device, 113 is a video camera 111 and a CRT device 11
2 and a video conference terminal equipped with a microphone speaker, 115
Is a line connecting the video conference terminal and the multipoint control unit.

FIG. 19 is a block diagram showing the inside of the communication control apparatus. In the figure, 110 is a multipoint communication control apparatus, and 102 is a system control section for controlling and managing the whole. Reference numeral 103 denotes a line interface unit (hereinafter referred to as a line I / F unit) that manages an interface with a line connecting each video conference terminal. 104-1 ~ 1
Reference numeral 04-n is a multiplexing unit that multiplexes audio data, video data, and general-purpose data transmitted to each of the video conference terminals, and 105-1 to 105-n are multiplex data received from each video conference terminal as audio data. , A separation unit that separates video data and general-purpose data.

Reference numeral 2 is a multiplexer 104 that synthesizes the audio data separated by the separators 105-1 to 105-n.
-1 to 104-n, an audio addition / distribution unit, and 7 is 1 of the video data separated by the separation units 105-1 to 105-n.
A video switching / distributing unit that selects one of the two and sends it to the multiplexing units 104-1 to 104-n, and 1 is a speaker determining unit that determines a speaker based on voices from a plurality of video conference terminals.

FIG. 20 is a block diagram showing the structure of the voice addition / distribution unit 2. In the figure, 20 is a voice management unit that manages the entire voice addition and distribution unit 2, and 21 is a voice synthesis unit that adds the voice data sent from each of the separation units 105-1 to 105-n. 22-1 to 22-
n is a speech synthesis unit 21 that converts the non-linear voice data from each of the separation units 105-1 to 105-n into linear voice data.
To the multiplex units 104-1 to 104-1 to convert the linear voice data synthesized by the voice synthesis unit 21 into nonlinear voice data.
104-n is a voice encoding / decoding unit. Reference numeral 23 is an n-bit voice detection flag provided in the voice management unit 20 and each bit corresponds to each video conference terminal.

Next, the operation will be described. Among the video conference terminals connected to the line I / F unit 103 via the line,
The video conference terminal (parent terminal) that intends to hold a conference instructs the multipoint communication control device 110 which video conference terminal and the multipoint video conference to be held.

Upon receiving the request, the system control unit 102 gives an instruction to the line I / F unit 103 to call the designated video conference terminal. When the line between each called video conference terminal and the multipoint communication control device 110 is connected, transmission and reception of multiplexed data in which voice data, video data, and general-purpose data are multiplexed is performed, and the multipoint video conference is performed. Is started.

[0008] The multiplexed data transmitted from each video conference terminal is received by the line I / F unit 103 and then passed to the corresponding demultiplexing units 105-1 to 105-n,
Here, it is separated into audio data and video data. The audio data is sent to the audio addition / distribution unit 2 and the video data is sent to the video switching / distribution unit 3.
, Respectively. In the voice addition / distribution unit 2, the voice data from each of the separation units 105-1 to 105-n is received by the corresponding voice encoding / decoding unit 22-1 to 22-n,
The non-linear data is converted to linear data and sent to the speech synthesis unit 21. The voice synthesizing unit 21 adds the voice data converted into linear data by the voice encoding / decoding units 22-1 to 22-n, respectively.

After that, voice data obtained by subtracting one's own voice from the added voice data is generated and passed to each of the corresponding voice encoding / decoding units 22-1 to 22-n. Each of the voice coding / decoding units 22-1 to 22-n converts the received voice data from linear data to non-linear data and sends it to the corresponding multiplexing units 104-1 to 104-n. Further, the voice synthesizing unit 21 turns on the voice detection flag corresponding to the video conference terminal in which the voice is detected, and repeats the operation of clearing all n bits for each predetermined time unit (hereinafter, referred to as voice detection period). .

The video data separated by each of the separation units 105-1 to 105-n is sent to the video switching distribution unit 3, and the video switching distribution unit 3 receives the video data according to the instruction of the speaker judging unit 2. One of the multiplex units 10
The video data is switched so as to be distributed to 4-1 to 104-n.

Next, FIG. 21 shows, for example, Japanese Patent Publication No. 3-6570.
It is a block diagram which shows the conventional speaker determination part shown by the 8th publication. In the figure, reference numeral 1 is a speaker determination unit which is provided in a multipoint communication control device used in a multipoint video conference system or the like and which determines a speaker. 2 is an audio addition and distribution unit that adds audio from a plurality of connected video conference terminals and distributes the audio to each video conference terminal, and 3 distributes video from a plurality of connected video conference terminals to other video conference terminals It is a video switching distribution unit. Reference numeral 4 is a control processing unit that manages execution of each processing unit in the speaker determination unit 1 and performs speaker determination, 5 is a speaker determination parameter that stores a speaker determination parameter for performing speaker determination, and 6 is A voice count unit that counts the number of voice detections from the voice detection flag 23 of the voice addition and distribution unit 2 according to an instruction from the control processing unit 4 to detect the presence or absence of voice, and 7 is a voice detection result (count value of the voice count unit 6 ) Is stored in the voice count memory, and 8 is a voice detection result (count value) stored in the voice count memory 7 according to an instruction from the control processing unit 4 less than a threshold value recorded in advance in an effective voice count. If so, an effective voice determination unit that clears the voice detection result (count value) of the corresponding voice count memory 7, and 9 is stored in the voice count memory 7 according to an instruction from the control processing unit 4. The one having the largest voice detection result (count value) than the detected voice detection result is searched, the terminal is determined as the speaker, and the video of the determined speaker is distributed to the video switching distribution unit 3 to other terminals. It is a maximum voice count determination unit for instructing to.

Further, FIG. 22 shows a configuration example of the speaker determination parameter 5 and the voice count memory 7. Reference numeral 5a is a speaker determination counter for designating a period for speaker determination, 5b is a voice detection time for designating a voice detection time interval, 5c.
Is a valid voice counter for designating a threshold value for determining the voice detection result as valid, and 7a to 7n are the voice detection results of the connected video conference terminals from the voice detection flag 23 of the voice addition and distribution unit 2. Judgment is 1 when there is voice detection
Is a voice counter for storing the result of addition. As an example of the speaker determination method, the speaker determination counter 5a is set to a value that is an integral multiple when the voice detection time 5b is 1.

In the example shown in FIG. 22, the voice detection time 5b is set to 40 ms, and when this 40 ms is set to 1, the speaker determination counter is set to a value 50 times that value. The value of the speaker determination counter indicates the speaker determination period by a counter value, and 40 ms is 5 in this example.
2 seconds when 0 pieces are collected is defined as a speaker determination period. This speaker determination period (2 seconds) is for determining the speaker in this time unit, and means that the same speaker is displayed for at least 2 seconds even if the speaker changes rapidly. . FIG. 23 is a diagram showing the relationship between the speaker determination period and the voice detection time. In the figure, C1, C2, C3 and C4 indicate respective speaker determination periods (2 seconds). Numbers 1 to 50 indicate that 50 voice detection times (40 ms) exist within one speaker determination period (2 seconds).

If the effective voice counter 5c shown in FIG. 22 detects voice detection more than 50 times within the number of voice detections within the speaker determination period, it can be a speaker candidate within the speaker determination period. Set a threshold for that. In the example shown in the figure, a value of 30 is set. Therefore, when the voice is detected 30 times or more out of the 50 times of voice detection, the speaker of the terminal is set as the candidate for screen switching. From the meanings of the speaker determination counter and the valid voice counter described above, the value of each counter is "speaker determination counter 5a ≧ valid voice counter 5c.
It is assumed that it is set under the condition of ">0".

Next, the operation will be described. Speaker determination unit 1
In, the control processing unit 4 instructs the voice counting unit 6 to clear the voice counters 7a to 7n of the voice counting memory 7 to 0 before performing the speaker determination.

Next, the control processing unit 4 uses the speaker determination parameter 5
After waiting until the voice detection time (40 ms) has passed, the voice counting unit 6 is instructed to perform voice detection. During this voice detection time, the voice synthesizing unit of the voice adding / distributing unit 2 inputs voices from each video conference terminal and adds and distributes the voices. Further, the voice synthesizing unit turns on the voice detection flags 23 corresponding to the terminals in which voices are detected during the voice detection time. For example, when voice is input from the first and second video conference terminals, the first and second flags of the voice detection flag 23 are turned on.

The voice counting unit 6 detects the presence or absence of voice detection for each video conference terminal from the voice detection flag 23 of the voice addition and distribution unit 2, and when there is voice detection, the voice count memory 7
The voice counters 7a to 7n of the corresponding video conference terminal are incremented by one. For example, when the first and second bits of the voice detection flag 23 are on, 1 is added to the voice counter 7a of the terminal 1 and the voice counter 7b of the terminal 2. The voice detection flag 23 of the voice addition / distribution unit 2 outputs the value of the flag to the speaker determination unit 1 and then clears all n bits of the voice detection flag, and a new voice is input at the next voice detection time. The corresponding flag of the voice detection flag 23 is turned on again for the terminal.

The control processing unit 4 repeats the above processing until the value of the speaker determination counter 5a of the speaker determination parameter is reached. In this example, since the count value of the speaker determination counter 5a is 50, the above process is repeated 50 times. On the other hand, the voice detection flag 23 of the voice addition and distribution unit 2 also repeats detection and clearing 50 times. In this way, the voice counters 7a to 7n record the count value of how many times each voice is detected within the speaker determination period (2 seconds). Each value of the voice counter of the voice count memory shown in the figure shows each counter value when the speaker determination period (2 seconds) has elapsed, and in this example, terminal 1 to 45 times, terminal 2 to 15
, And 40 times from the terminal n, respectively.

Next, the control processing unit 4 causes the effective voice determining unit 8 to set the value (30) of the effective voice counter 5c of the speaker determination parameter.
Is instructed and an instruction to determine valid voice is given. The valid voice count determination unit 8 is a voice counter 7a of the voice count memory 7.
If ~ 7n is less than the valid sound counter value (30), the corresponding sound counters 7a to 7a are cleared to 0. At this time point, the terminal 2 shown in the figure has a voice counter value smaller than 30, and is cleared to 0.

Next, the control processing unit 4 instructs the maximum voice count determination unit 9 to determine the speaker. Maximum voice count determination unit 9
Judges the speaker with the largest sound count value among the sound counters 7a to 7n of the sound count memory 7 as the speaker, and distributes the video of the video conference terminal corresponding to the video switching / distributing unit 3 to other video conference terminals. Instruct them to do so. In this example, since the terminal 1 has a larger count value than the terminal n, the terminal 1 is determined as the speaker.

When there are two or more voices having the highest voice count value, the speaker with the earliest connected terminal is determined as the speaker. Further, when all the audio counters 7a to 7n are 0, it is assumed that there is no speaker and the video switching distribution unit 3 is not notified. When the speaker is not notified to the video switching distribution unit 3, the video switching distribution unit 3 continuously displays the speaker who has been displayed before.

By repeating the above operation, the speaker determination method is realized by using the value of the speaker determination counter 5a of the speaker determination parameter 5 as the video switching interval by the speaker determination.

[0023]

Since the conventional speaker determination method is configured as described above, the video switching by the speaker determination is performed for the speaker with the maximum number of voice detections per speaker determination period. Therefore, there is a problem that video switching occurs every time the speaker determination period occurs, and video switching is unnatural.

The present invention has been made in order to solve the above-mentioned problems, and prioritizes the previous speaker in the video switching by the speaker determination and prioritizes the voice detection in the rear part of the speaker determination period. The purpose of the present invention is to obtain a speaker determination method that makes video switching as natural as possible by changing the speaker determination period.

[0025]

A speaker determination method according to a first aspect of the present invention is a speaker determination means for storing a terminal determined to be a speaker in a previous speaker determination period as a previous terminal,
In the present speaker determination period, a plurality of candidate terminals that are candidates for speaker terminals are selected based on a predetermined criterion, and when the previous terminal stored by the previous speaker storage means is included in the selected candidate terminals, the above The present invention is characterized in that a speaker determination unit that uses the previous terminal as the current speaker terminal is provided.

The speaker determination method according to the second invention is based on the number of voice detection periods determined to be voice among a plurality of voice detection periods in the speaker determination period as a predetermined criterion for selecting candidate terminals. It is characterized by using the required voice frequency.

The speaker determination method according to the third aspect of the invention is based on the position of the voice detection period determined to be voice among a plurality of voice detection periods in the speaker determination period as a predetermined criterion for selecting candidate terminals. It is characterized by using the required voiced position.

The speaker determination method according to the fourth aspect of the invention is based on the number of voice detection periods determined as voice among a plurality of voice detection periods in the speaker determination period as a predetermined criterion for selecting candidate terminals. It is characterized in that both the required voice frequency and the voice position in the speaker determination period of the voice detection period determined to be voice are used.

In the speaker determination method according to the fifth aspect of the present invention, the values of the parameters used for the above-mentioned predetermined criterion held by the previous terminal and other candidate terminals are compared, and when the difference is less than a predetermined threshold value. , The previous terminal is a speaker terminal.

In the speaker determination system according to the sixth aspect of the present invention, the speaker determination period is divided into zones having a plurality of voice detection periods, and the speaker terminal is determined based on a predetermined standard in units of zones. It is characterized by

In the speaker determination method according to the seventh aspect of the invention, the speaker is detected based on the position in the speaker determination period of the voice detection period determined to be voice among a plurality of voice detection periods in the speaker determination period. It is characterized by determining the person terminal.

In the speaker determination method according to the eighth aspect of the present invention, the difference between the terminal determined as the speaker in the previous speaker determination period and the terminal determined as the speaker in the current speaker determination period is It is characterized in that it comprises an interval changing means for changing an interval for judging a person.

The speaker determination method according to the ninth aspect of the present invention is as follows when the terminal determined as the speaker in the previous speaker determination period and the terminal determined as the speaker in the current speaker determination period are different. A speaker fixing means for delaying the start time of the speaker determination period for a predetermined time is provided.

A speaker determination system according to a tenth aspect of the invention comprises a speaker determination period storage means for storing the speaker determination period,
The interval changing means includes a speaker determination period changing means for changing the speaker determination period stored in the speaker determination period storage section.

The speaker determination system according to the eleventh aspect of the invention is characterized in that a period obtained by multiplying the speaker determination period by a predetermined multiplication factor is set as a next speaker determination period.

In the speaker determination method according to the twelfth invention, when the terminal determined as the speaker in the previous speaker determination period and the terminal determined as the speaker in the current speaker determination period are the same, The speaker determination period obtained by deleting a part of the speaker determination period is set as the next speaker determination period.

In the speaker determination method according to the thirteenth invention, a speaker determination parameter storage unit storing various parameters used for speaker determination and a parameter different from the parameter value stored in the speaker determination parameter storage unit. The default parameter storage unit that stores a parameter having a value of, the terminal that is determined as the speaker in the previous speaker determination period and the terminal that is determined as the speaker in the current speaker determination period are different from each other in the above speaker. It is characterized by further comprising a parameter changing means for changing the value of the parameter stored in the determination parameter storage unit according to the value of the parameter stored in the default parameter storage unit.

[0038]

In the speaker determining method according to the first aspect of the invention, the speaker determining means stores the terminal determined as the speaker in the previous speaker determining period as the previous terminal as the previous terminal. If multiple candidate terminals are selected from the speaker determination period of this time according to the predetermined criteria and the previous terminal is included, the terminal selected last time is given priority because the previous terminal is the speaker terminal of this time. It becomes the speaker terminal. Therefore, when the speaker determination period elapses, the possibility that the same terminal continues as a speaker increases, and the inconvenience that the speaker is frequently switched for each speaker determination period is eliminated.

The speaker determination system according to the second invention is characterized in that the voice frequency is used as a criterion for selecting candidate terminals, and some candidate terminals are selected in descending order of voice frequency. If the selected candidate terminal includes the previous terminal, this is set as the speaker terminal.

The speaker determination method according to the third aspect of the invention is characterized in that the position within the speaker determination period of the voice detection period determined as voiced is used as a criterion for selecting candidate terminals. A terminal that is detected as having a sound behind in the determination period is set as a candidate terminal, and if the terminal is included in the candidate terminals selected by the previous terminal, the previous terminal is set as the speaker terminal. It is highly possible that a terminal having a voice detection period which is voiced after the speaker determination period continues to output a voice continuously in the next speaker determination period.
Therefore, it is possible to prioritize the previous speaker and select a terminal that has a high possibility of becoming the next speaker as the speaker terminal, increasing the possibility that the same terminal will continue as the speaker. The inconvenience that the speaker is frequently switched for each period disappears.

The speaker determination system according to the fourth aspect is characterized in that the speaker is determined by using both the voice frequency and the voice position. In this case, the voice frequency may be prioritized, some candidate terminals may be selected, and the speaker terminal may be selected from the voice positions of the selected candidate terminals. Alternatively, conversely, a candidate terminal may be selected from the voiced position, and the frequency of the voice with respect to the selected terminal may be compared to select the speaker terminal.

In the speaker determination method according to the fifth aspect of the present invention, when the candidate terminal includes the previous terminal, the candidate terminal has the largest parameter value used for the predetermined reference and the previous terminal. If the difference is less than or equal to a predetermined threshold, the previous terminal is set as the speaker terminal, so even if the previous terminal is included in the candidate terminals, the If there is a difference, the speaker terminal is the most likely candidate terminal. On the contrary, when the difference between the candidate terminal having the highest value among the candidate terminals and the previous terminal is less than or equal to the threshold value, the previous terminal is prioritized as the speaker terminal.

The speaker determination system according to the sixth aspect of the invention is characterized in that the speaker determination period is divided into a plurality of zones and the voiced position is determined in units of the zones. Although it is possible to use the voiced position of the voice detection period, by using the zone having a plurality of voice detection periods in this way, for example, when noise is generated or a state where voice is generated for a very short time is generated. It can be ignored without determining the voiced position. That is, it is possible to more reliably detect whether or not the speaker is continuously outputting a voice signal by determining whether or not there is sound for a certain period of time called a zone.

The speaker determination system according to the seventh invention is characterized in that the previous speaker storage means in the above-mentioned invention is not provided and the speaker terminal is simply determined based on the voiced position. There is. Using a terminal that has a voice detection period determined to be voiced later than the speaker determination period as a speaker terminal may continue to be the speaker terminal during the next speaker determination period. It is to select a high terminal, and by selecting such a terminal, it is possible to prevent the speaker terminal from being switched at each speaker determination period.

In the speaker judging system according to the eighth aspect of the invention, the terminal judged as the speaker in the previous speaker judging period and the terminal judged as the speaker in the present speaker judging period are different from each other. The interval for determining the speaker such as the speaker determination period is changed. That is, when the speaker terminal of the previous time and the speaker terminal of this time are different, the speaker terminal of this time may be switched immediately in the next speaker determination period, so the interval is made longer. On the other hand, when the previous terminal and the current terminal are the same, the current terminal has been continuously determined as the speaker terminal from the previous time, and therefore the speaker determination period is set to the conventional value or shortened. In this way, it is possible to prevent the terminal determined by changing the interval due to the difference between the previous terminal and the current terminal from frequently switching for each speaker determination period.

In the speaker determination method according to the ninth aspect of the invention, the start of the next speaker determination period is shifted backward in order to change the interval, and a predetermined fixed time is inserted, so that only the fixed time is inserted. The determined speaker terminal will continue to be selected for a longer period than usual.

In the speaker determination method according to the tenth aspect of the invention, the stored speaker determination period is rewritten, and a new speaker determination period is used by rewriting this speaker determination period. The next speaker decision is made.

In the speaker determination method according to the eleventh invention, a new speaker determination period is calculated by multiplying the speaker determination period by a predetermined multiplication factor in order to change the speaker determination interval. is there.

In the speaker determination method according to the twelfth invention, if the previous terminal and the current terminal are the same, the next speaker determination period may be shortened, and the next speaker determination period This is to remove a part.

In the speaker determination system according to the thirteenth invention, various parameters used for speaker determination are stored in the speaker determination parameter storage section, and the speaker terminal of the previous time and the speaker terminal of this time are distinguished from each other. This is to rewrite the speaker determination parameter. By rewriting this parameter, not only the interval for determining the speaker as described above is changed, but also other parameters such as the change of the threshold value or the change of the sampling value are performed. It is also possible to change the parameters.

[0051]

EXAMPLES Example 1. Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of the present invention. In the figure, 1 is a speaker determination unit, 2 is a voice addition / distribution unit, 3 is a video switching distribution unit, 4 is a control processing unit, 5 is a speaker determination parameter, 6 is a voice counting unit, 7 is a voice counting memory, and 8 is a voice counting memory. Is an effective voice determination unit. They are,
The detailed description is omitted because it is the same as or equivalent to that of the conventional example.

In response to an instruction from the control processing unit 10, numeral 10 obtains the upper two terminals having a larger voice detection result (count value) than the voice detection result stored in the voice count memory 7, and outputs the voices of the two higher terminals. If the difference between the counts is less than the speaker determination threshold value and one of them is the previous speaker, the terminal is set as the current speaker, and in the other cases, the top 1
The top 2 voice count determination unit that sets the terminal of the highest rank as the speaker of this time, and the reference numeral 11 is a previous speaker storage memory that stores the terminal determined as the speaker by the top 2 voice count determination unit 10.

FIG. 2 shows a configuration example of the speaker determination parameter 5 and the voice count memory 7. Reference numeral 5a is a speaker determination counter for designating a period for speaker determination, 5b is a voice detection time for designating a time interval of voice detection, 5c.
Is a valid voice counter for designating a threshold value for determining that the voice detection result is valid, and 5d is a zone voice counter for designating one zone period when the speaker determination period is divided into a plurality of zones. 5e, if the difference when obtaining the top two terminals with a large amount of voice detection is a certain value or more,
The speaker identification counter 5f for unconditionally specifying the threshold value for determining the top-ranked terminal as the speaker is a speaker determination process when the video is switched by the speaker determination. It is a fixed speaker time to delay.

Reference numerals 77a to 77n are memories for storing the voice detection result for each connected video conference terminal.
0 to 77n0 are total voice counters for storing the result of adding 1 when there is voice detection, 77a1 to 77a1
Reference numerals 77am to 77n1 to 77nm are zone voice counters for storing the result of adding 1 with 1 when voice is detected for each zone.

Since the configuration examples of the speaker determination parameter 5 and the voice count memory 7 shown in FIG. 2 include all those used in the embodiments described later, the details will be described in each embodiment.

As an example of the speaker determination method, the speaker determination counter 5a and the zone voice counter 5d are set to "speaker determination counter 5a = zone voice counter 5d × integer", and the speaker determination counter 5d and the valid voice counter are used. 5c and the speaker identification counter 5e are set under the condition of "speaker determination counter 5a≥valid voice counter 5c≥speaker identification counter 5e>0".

Next, the operation will be described. Here, FIG. 3 is a flowchart showing the flow of the operation. In step ST1, the control processing unit 4 instructs the voice counting unit 6 to clear the total voice counters 77a0 to 77n0 of the voice counting memory 7 to 0 before performing the speaker determination.
Next, in step ST2, the control processing unit 4 waits until the voice detection time 5b of the speaker determination parameter 5 elapses.
When the voice counting unit 6 is instructed to perform the voice counting process in step ST3, the voice counting unit 6 detects from the voice addition and distribution unit 2 whether or not voice detection is made for each video conference terminal. Total audio counters 77a0-77n of the corresponding video conference terminal
Increment 0 by 1. In step ST4, the control processing unit 4
Repeats the processing of steps ST2 to ST4 until the count value of the speaker determination counter 5a of the speaker determination parameter 5 is reached. FIG. 4 is a diagram showing an example of the value of the total voice counter when this repeating process is completed. For example, the total voice counter of the terminal 1 is 40. Similarly, 43 for terminal 2 and 3 for terminal 3.
1 and 20 for terminal n.

Next, in step ST5, the control processing unit 4 notifies the valid voice determination unit 8 of the value of the valid voice counter 5c of the speaker determination parameter 5, and instructs the valid voice determination process. If the total voice counters 77a0 to 77n0 of the voice count memory 7 are less than the value of the valid voice counter 5c, the valid voice determination unit 8 clears the corresponding total voice counters 77a0 to 77n0. In this example, since the valid voice counter 5c is set to 30, in FIG. 4, a value smaller than 30 is cleared to 0. Therefore, the total voice counter 77 of the terminal n
n0 is cleared to 0. The reason why the terminal having the value smaller than the value indicated by the effective voice counter is eliminated by using the effective voice counter is to keep the number of times for judging the speaker as a certain number or more. If the value of this voice counter is increased, many states that there is no sound will be detected, and if the value of this effective voice counter is decreased, many states that there is sound will be detected.

Next, in step ST6, the control processing unit 4 notifies the upper 2 voice count processing unit 10 of the threshold value of the speaker identification counter 5e of the speaker determination parameter 5, and instructs the upper 2 voice count determination process. . The upper 2 audio count processing unit 10 seeks the upper 2 video conference terminals having a larger audio count value than the total audio counters 77a0 to 77n0 of the audio count memory 7, and the difference between the audio count values is the threshold of the speaker identification counter 5e. When the value is less than the value, if the previous speaker video conference terminal stored in the speaker storage memory 11 last time is the second video conference terminal, the second video conference terminal is determined to be the speaker, and if the previous speaker video conference terminal is the first speaker otherwise. The video conference terminal is determined to be the speaker of this time, and the video conference terminal of the speaker is stored in the speaker storage memory 11 last time.

In the example shown in FIG. 4, the top two video conference terminals are terminal 1 and terminal 2. And the difference is 4
Since 3-40 = 3, it is smaller than the threshold value (10) of the speaker identification counter 5e shown in FIG. Therefore, if the terminal 1 was selected in the previous speaker determination period and is the terminal stored in the speaker storage memory 11 last time, the terminal 1 is determined again as the speaker in the current speaker determination period. .

When the values of the total audio counters 77a0 to 77n0 of the first or second terminal and the third and subsequent terminals are the same, the connected video conference terminals in the youngest order are the first and second terminals. When the total voice counts 77a0 to 77n0 are all 0 for the terminal, the last speaker is determined to be the video conference terminal.

Next, in step ST7, when the video conference terminal determined to be the current speaker and the video conference terminal of the previous speaker are not the same, the upper 2 audio count processing unit 10 sends the video switching distributor 3 to the speaker. It is instructed to distribute the video of the video conference terminal determined to be to the other video conference terminals.

The operations of steps ST1 to ST7 are repeated, and the speaker determination counter 5a of the speaker determination parameter 5 is repeated.
A speaker determination method is realized by using as a video switching interval by speaker determination.

As described above, in this embodiment, in the speaker determination system for switching the video of a plurality of video conference participating terminals in the multipoint communication control device or the like, the speaker determination configured on the speaker determination unit is performed. A speaker determination parameter that stores a parameter, a control processing unit that issues a speaker determination instruction based on the parameter, a voice counting unit that counts voice detection results from the video conference terminal, and a voice that stores the voice count result. With a count memory and a valid voice counter (voice threshold), a valid voice determination unit that clears a voice count result below the threshold from the voice count memory,
The previous speaker storage memory that stores the previous speaker and the top two video conference terminals that have a higher voice count value than the voice count memory are obtained, and the difference between the voice counts is less than the speaker specific counter (speaker specific threshold). When either one is the previous speaker, the video conference terminal is used as the speaker, and in the other cases, the top two audio count determination unit using the first video conference terminal terminal as the speaker is provided. The speaker determination method was explained.

In the speaker determination system according to this embodiment, the previous speaker storage memory for storing the previous speaker and the top two video conference terminals having the highest voice detection frequency are obtained, and the difference between the voice detection frequencies It is characterized in that a higher-ranked two-voice count determination unit for determining the last-time speaker's video-conference terminal or the video-conference terminal with the first highest audio detection amount as a speaker is provided. The upper 2 audio count determination unit obtains the difference in the amount of audio detection between the upper 2 video conference terminals with the higher audio detection amount, and when the difference is less than the speaker identification threshold, the previous time stored in the speaker storage memory. Since the comparison with the speaker is performed, when the difference in the voice detection amount between the terminal with the highest voice detection amount and the terminal with the second highest voice detection amount is small, the terminal with the second highest voice detection amount can be the speaker last time. For example, a speaker determination method that does not switch the video is realized because it is determined as the speaker.

Example 2. In the first embodiment, the upper two video conference terminals having the larger audio count value are selected as the candidate terminals in the upper two audio count generation unit 10. However, the selection is not limited to the case of selecting the upper two terminals, and for example, the upper three. A predetermined plurality of terminals may be used as candidate terminals as in the case of selecting one terminal or upper four terminals.

For example, in FIG. 4, when the top three video conference terminals are used as the candidate terminals, the three terminals 1, 2, 3 are selected as the candidate terminals. Similarly, the value 10 of the speaker identification counter 5e is used to determine whether or not the difference is within the threshold value. In this example, terminal 3 shows the value 31, terminal 2 shows the value 43, and the difference is 1
Since the terminal 3 is larger than 0, the terminal 3 stores the speaker storage memory 1
Even the previous speaker terminal stored in 1 is not selected as the current speaker terminal. In this way, even if a terminal separated by more than the threshold value set in the speaker identification counter 5e is stored in the speaker storage memory last time, the speaker determination period of this time is not set as the speaker terminal of this time. This is because a terminal more distant by a threshold value or more is removed from the one showing the maximum total voice count value in to select a more suitable speaker.

Example 3. In the above-described first and second embodiments, when the voice counter value between the candidate terminals is not equal to or more than the predetermined threshold value by using the speaker identification counter 5e, the last talk is performed in the candidate terminal. Although the case where the terminal stored in the speaker storage memory is included is not used as the speaker terminal this time, the threshold value of the speaker storage counter 5e is not used, and the plurality of terminals simply selected as candidate terminals are not used. When the terminal includes the terminal stored in the speaker storage memory last time, the terminal may be the speaker terminal of this time.

Example 4. FIG. 5 is a block diagram showing another embodiment of the present invention. In the figure, the same symbols as those in FIG. 1 are the same or corresponding parts, and the description thereof will be omitted.

Reference numeral 6 indicates the presence or absence of voice from the voice addition / distribution unit 2 in response to an instruction from the control processing unit 4, and increments the value of each of the total voice counters 77a0 to 77n0 of the voice count memory 7 by adding 1 thereto. At the same time, the zone voice counters 77a1-7 designated by the control processing unit 4
7am, ~, 77n 1 ~ 77nm to increase the value by adding 1 to the voice count unit, 12 is the zone voice counter 77am ~ 77nm from the zone sound counter located at the rearmost in the voice count memory 7 according to an instruction from the control processing unit 4. Counters 77a1 to 77am, ..., 77n1 to 77nm seek a video conference terminal that is not 0, and in the same zone,
When there are two or more video conference terminals, the upper two audio count determination unit 10 is instructed to determine the speaker, and when there is only one video conference terminal, the posterior zone audio whose speaker is the video conference terminal is used. It is a judgment unit.

Next, FIG. 6 is a diagram for explaining the zones in this embodiment. In FIG. 6, C1, C2,
C3 and C4 are speaker determination periods. Further, Z1, Z2, Z3, Z4 and Z5 in the speaker determination period are zones. In this example, the speaker determination period (2 seconds) is divided into five zones. Therefore, one zone is 40
It is 0 ms, and there are 10 voice detection times each. In this way, the zone is created by dividing the speaker determination period and has a plurality of voice detection times.

Next, the operation will be described. Here, FIG.
Is a flowchart showing the flow of the operation. In step ST1, the control processing unit 4 instructs the voice counting unit 6 before performing the speaker determination, and the total voice counters 77a0 to 77n0 and the zone voice counters 77a1 to 77am, ..., 77n1 to 77 of the voice count memory 7 are instructed.
Clear all nm.

Next, in step ST2, the control processing unit 4 waits until the voice detection time 5b of the speaker determination parameter 5 elapses. In step ST3, the voice counting unit 6
Notify the zone position to and instruct voice count processing.

The voice counting unit 6 detects the presence or absence of voice detection for each video conference terminal from the voice addition and distribution unit 2. When voice is detected, the total voice counter 77a0 to 77n0 of the corresponding video conference terminal in the voice count memory 7 is detected. And zone voice counters 77a1 to 77am, ...
+1 for 77n1 to 77n.

For example, in the zone Z1 of the speaker determination period C1 shown in FIG. 6, when the voice is detected five times,
The zone 1 voice counter 77a1 of the terminal 1 shown in FIG.
Will indicate the value. Similarly, when the voice is detected four times in the zone Z2, the value of the zone 2 voice counter 77a2 becomes 4 as shown in FIG.

In step ST4, the control processing unit 4 repeats the processing of steps ST2 to ST4 until the number of voice count processing reaches the speaker determination counter 5a of the speaker determination parameter 5. Note that the zone position is set to an initial value of 1, and in step ST4, the control processing unit 4 changes the zone position by setting it to +1 every time the number of voice count processes reaches the zone voice counter 5d of the speaker determination parameter 5. To do. At the time when the process of step ST4 is completed, the repeating process is repeated 50 times. The value of the total sound counter 77a0 shown in FIG.
It is a value obtained by adding the values up to 7 am.

FIG. 8 is a diagram showing an example of the case where the processing in step ST4 is completed, and is a diagram showing an example of the total voice counter values of terminals 1 to 4 and the voice counter values of zones 1 to 5. is there.

Next, in step ST5, the control processing unit 4 notifies the effective voice determination unit 8 of the value of the effective voice counter 5c of the speaker determination parameter 5, and instructs the effective voice determination process. If the total voice counters 77a0 to 77n0 of the voice count memory 7 are less than the valid voice counter 5c, the valid voice determination unit 8 applies the corresponding total voice counter 7.
7a0 to 77n0 and zone voice counters 77a1 to 77
Am, ~, 77n1-77nm is cleared to 0. In the case of the example shown in FIG. 8, since the total voice counter of the terminal 4 shows a value smaller than the effective voice counter value (30), the total voice counter and the zone voice counter are 0.
Cleared.

Next, in step ST8, the control processing unit 4 causes the posterior zone voice determination unit 12 to determine the speaker determination parameter 5
The threshold value of the speaker identification counter 5e is notified, and the subsequent zone voice determination processing is instructed. Rear zone voice determination unit 12
Is obtained for each zone until one or more video conference terminals whose zone voice counters 77a1 to 77am, ..., 77n1 to 77nm are not 0 are found in order from the zone voice counters 77am to 77nm located at the rearmost of the voice count memory 7. In the case of the example shown in FIG. 8, since terminals 1 and 3 in zone 5 have values other than 0, terminal 1
Then, the terminal 3 is raised to the candidate terminal which becomes the speaker terminal.

Next, in step ST9, the posterior zone audio determination unit 12 determines that the speaker is undecided when the two or more video conference terminals are found, and determines the upper two audio count determination unit 1.
The threshold value of the speaker identification counter 5e of the speaker determination parameter 5 is notified to 0, and the upper 2 voice count determination process is instructed. Steps ST6 and ST7 perform the same operations as those in the first embodiment. If one video conference terminal is found, the video conference terminal is determined to be the speaker, and the previous video conference terminal of the speaker is stored in the speaker storage memory 11.
If the video conference terminal is not found, the video conference terminal of the speaker previously stored in the speaker storage memory 11 is determined to be the speaker.

In the example shown in FIG. 8, terminal 1 and terminal 3
Has been raised to the candidate terminal and the difference in the total voice counter of this candidate terminal is smaller than the threshold value (10) set in the speaker identification counter 5e. When the terminal is stored in the storage memory 11, the terminal stored in the speaker storage memory last time is determined as the speaker terminal of this time.

The above-described operations of steps ST1 to ST7 are repeated, and the speaker determination method is realized by using the speaker determination counter 5a of the speaker determination parameter 5 as the video switching interval based on the speaker determination.

As described above, in this embodiment, the voice is detected from the video conference terminal by the voice counting section which counts each of the zones obtained by dividing the speaker determination period into a plurality of zones, and the zone which is the furthest behind becomes the speaker. It is characterized in that it is provided with a rear zone deciding unit which searches for a video conference terminal.

The voice counting unit counts and stores the voice detection result from the video conference terminal for each zone, and the rear zone voice determining unit can search the video conference room to be the speaker from the zone located at the rearmost. Therefore, it is possible to realize the speaker determination method in which the video conference terminal whose voice is detected in the rear part is determined as the speaker in the speaker determination period and the video is switched.

According to this embodiment, the voice counting unit for counting the detection result of the voice from the video conference terminal for each zone and the rear zone voice judging unit are added to the speaker judging unit. It is possible to obtain a speaker determination method in which the video conference terminal detected at the rear part is determined as a speaker in the speaker determination period and video switching is performed.

Example 5. In the above embodiment, the case where the previous speaker storage memory 11 stores the previous speaker terminal has been described, but it does not matter if the previous speaker storage memory does not exist as shown in FIG. In the example shown in FIG. 9, the previous speaker storage memory 11 is deleted from FIG. 5 described in the fourth embodiment. Therefore, in the example shown in FIG. 9, first, the posterior zone audio determination unit 12 searches for the video conference terminal that is the speaker from the zone located furthest behind, and the upper two audio count determination unit 10 searches for the posterior zone audio determination unit 12 When multiple terminals are listed as candidate terminals in
The speaker with the highest voice detection amount is selected to determine the speaker terminal of this time. In this case, the upper 2 voice counter determination unit does not perform determination using the threshold value of the speaker identification counter 5e as described above, but simply selects one having a larger voice detection amount as the speaker terminal.

Example 6. In the above embodiment, the top two
Although the case where the voice counter determination unit 10 finally selects the speaker terminal has been described, the case where the posterior zone voice determination unit 12 selects the final speaker terminal may be performed as illustrated in FIG. 10. Absent. In FIG. 10, the posterior zone voice determination unit 12 searches for a terminal as a speaker from the zone located furthest behind and selects the terminal as a speaker terminal. If there are multiple terminals in the same zone, the terminal with the smaller number is selected as the speaker terminal.

Example 7. In the sixth embodiment, the case where the rear zone voice determination unit 12 determines the terminal to be the speaker from the zone located furthest to the rear has been described.
As shown in FIG. 1, when the previous speaker terminal is stored in the previous speaker storage memory 11 and there are a plurality of terminals in the zone located furthest in the posterior zone voice determination unit, the plurality of terminals are used. If the previous speaker terminal stored in the previous speaker storage memory 11 is included in, the previous speaker terminal may be set as the current speaker terminal.

Example 8. In the sixth and seventh embodiments described above, the case where the posterior zone voice determination unit 12 determines the terminal which is the speaker from the zone located furthest to the rear has been described. Not limited to this, the terminal to be the speaker may be determined from a plurality of zones including the zone located most rearward. For example, a terminal may be set as a candidate terminal for a terminal that is determined to have sound in the zone located at the rearmost and the zone located next thereto, that is, the two zones from the zone located at the back.

Example 9. In the above-described fourth embodiment, the case where the higher-order 2 voice count determination unit 10 selects one terminal when the posterior zone voice determination unit 12 detects a plurality of terminals from the rearmost zone has been described. In this embodiment, as shown in FIG.
The voice count determination unit 10 selects a plurality of candidate terminals first, and the posterior zone voice determination unit 12 talks about a terminal that is determined to be voiced in the zone located furthest out of the selected candidate terminals. The case of using the person terminal will be described. As described above, the upper 2 voice count determination unit 10 determines two upper terminals based on the voice detection amount. The posterior zone voice determination unit 12 selects, as a speaker terminal, a terminal which is determined to be sound in the zone located furthest from the two terminals. In FIG. 12, the previous speaker storage memory 1
Although the case where there is no 1 is shown, in FIG.
It has the previous speaker storage memory 11, and the posterior zone voice determination unit 12 preferentially selects the previous speaker terminal stored in the previous speaker storage memory 11 as the current speaker terminal. You may do so.

Example 10. FIG. 13 is a block diagram showing an embodiment of the present invention. In the figure, those denoted by the same reference numerals as those in FIG. 1 are the same or corresponding parts, and description thereof will be omitted.

Reference numeral 9 is a maximum voice count determination unit 9 which performs the same operation as that of the conventional one, and 13 is a maximum voice count determination unit 9
The video switching information memory that stores the speaker and the presence / absence of the video switching, 14 is the speaker fixed time when the video switching information stored in the video switching information memory 13 according to the instruction from the control processing unit 4 is the video switching. It is a speaker fixed determination unit that delays the speaker determination only.

Next, the operation will be described. Here in FIG.
Is a flowchart showing the flow of the operation. Since steps ST1 to ST5 and step ST7 perform the same operations as those of the first embodiment shown in FIG. 3, the description thereof will be omitted.

At step ST11, the control processing unit 4
Instructs the maximum voice count determination unit 9 to determine the speaker. The maximum audio count determination unit 9 determines that the speaker having the largest audio count value than the total audio counters 77a0 to 77n0 of the audio count memory 7 is the speaker, and the video of the video conference terminal corresponding to the video switching distribution unit 3 is changed to another TV. Instruct the conference terminals to distribute. At this time, the maximum audio count unit 9 refers to the previous speaker from the video switching information memory 13, and if the speaker is different from this time, the video is switched, and if the speaker is the same, the video is switched without switching the speaker and the video. The presence / absence is stored in the video switching information memory 13.

In step ST12, the control processing unit 4
Informs the speaker fixation determining unit 14 of the speaker fixation time 5f of the speaker determination parameter 5. The speaker fixing determination unit 14 refers to the presence / absence of video switching from the video switching information memory 13, and when the video switching is present, waits for the speaker determination fixed time 5f to elapse in step ST13.

FIG. 15 is a diagram showing an example of the case where the video fixing process ST13 is performed, and shows the case where the speaker terminal is switched from the terminal 1 to the terminal 2 as a result of the speaker determination in the speaker determination period C1. It is a figure. FIG. 15A shows the case of conventional terminal switching, and FIG. 15B shows the case of terminal switching in this embodiment. Figure 15
When the terminal switches every speaker determination period (2 seconds) as shown in (a), according to this embodiment, the speaker fixed time 5f shown in FIG. 2 is inserted every time the terminal switches. . Therefore, when the terminal 1 is switched to the terminal 2, the fixed time of the speaker (1 second) is longer than the normal time (2 seconds).
Is added time, that is, 3 seconds is the next time,
The phenomenon that the terminal is displayed for only 2 seconds disappears.
In this way, when the terminals are frequently switched in a short time, it gives an unpleasant feeling to the person watching the image. It is possible to eliminate discomfort.

When the video switching is performed by repeating the above-mentioned operations of steps ST1 to ST13, a value obtained by adding the value of the speaker determination counter 5a of the speaker determination parameter 5 and the value of the speaker determination fixed time 5f is set. , When there is no video switching,
The speaker determination method is realized by using the value of the speaker determination counter 5a of the speaker determination parameter 5 as the video switching time by the speaker determination.

That is, the speaker fixing determination section delays the start of speaker determination by creating a waiting state for a certain period of time when there is video switching depending on whether or not there is video switching stored in the video switching information memory. Because you can, after switching the video,
To realize a speaker determination method that does not switch the video immediately by speaker determination.

According to this embodiment, since the video switching information memory and the speaker fixing determination section are added to the speaker determination section, the video switching by the speaker determination can be performed immediately when the video switching is performed. Therefore, there is an effect that a speaker determination method capable of confirming the image of the speaker's video conference terminal can be obtained.

Example 11. FIG. 16 is a block diagram showing an embodiment of the present invention. In the figure, those given the same reference numerals as those in FIG. 1 and those given the same reference numerals as those in FIG. 13 are the same or corresponding parts, and a description thereof will be omitted.

Reference numeral 15 is a speaker determination parameter change processing unit for changing the speaker determination parameter 5 according to an instruction from the control processing unit 4 depending on the presence / absence of video switching stored in the video switching information memory 13, and 16 is a speaker determination parameter. Parameter change processing unit 1
5 is a speaker determination default parameter that stores different speaker determination parameters as defaults depending on whether video is switched. Here, the speaker determination default parameter has the same format as the speaker determination parameter of FIG. 2, and at least one value is different from the speaker determination parameter shown in FIG.

Next, the operation will be described. Here, FIG.
7 is a flowchart showing the flow of the operation. Since steps ST1 to ST7 perform the same operations as those shown in FIG. 14, the description thereof will be omitted.

At step ST12, the control processing unit 4
Instructs the speaker determination parameter change processing unit 15 to change the default parameters. The speaker determination parameter change processing unit 15 refers to the presence / absence of video switching from the video switching information memory 13, and when video switching is performed, the speaker determination default parameter 16 in step 14 is used as the speaker determination parameter when video switching is performed. Copy to parameter 5. When there is no video switching, the parameter of the speaker determination default parameter 16 when there is no video switching is copied to the speaker determination parameter 5 in step 15. As described above, the parameter stored in the speaker determination default parameter 16 has the same configuration as the speaker determination parameter stored in the speaker determination parameter 5, but the speaker determination parameter when video switching is performed is For example, in order to lengthen the video switching interval based on the speaker determination, it is assumed that at least the speaker determination counter 5a and the voice detection time 5b are set to be longer than when there is no video switching.

By repeating the above-mentioned operations of steps ST1 to ST15, the speaker determination method in which the video switching time by the speaker determination is longer is realized when the video is switched than when the video is not switched.

As described above, in the speaker determination method according to this embodiment, the video switching information memory for storing the presence / absence of the speaker by the speaker determination and the speaker determination default parameter for storing the default value of the speaker determination parameter. And a speaker determination parameter change processing unit for copying a default parameter having a different speaker determination period depending on the presence or absence of a speaker to the speaker determination parameter.

The speaker determination change processing section determines the default parameter having a long speaker determination period stored in the speaker determination default parameter when there is a speaker stored in the video switching information memory, to determine the speaker. Since it is copied to the parameter, the video switching by the speaker determination can be delayed, and the default parameter with a short speaker determination period stored in the speaker determination default parameter when there is no speaker is copied to the speaker determination parameter. To realize a speaker determination method that can speed up video switching by speaker determination.

According to this embodiment, the video switching information memory, the speaker determination default parameter, and the speaker determination parameter change processing unit are added to the speaker determination unit. The video switching based on the speaker determination is slow, and the video switching based on the speaker determination can be performed quickly when the video switching is not performed. Therefore, a speaker determination method that can smoothly switch the video of the speaker's video conference terminal can be obtained. .

Example 12. In Example 10 above,
Although the case where the speaker determination period is shifted by a fixed time has been described, the speaker determination period may be multiplied by one or more to set the next speaker determination period longer. For example, when 2 is used as the scaling factor, the next speaker determination period is 4 seconds. In this case, the speaker determination parameters shown in FIG. 2 are each doubled. Alternatively, the magnification may be 1 or less. For example, it may be 0.5. In this case, the speaker determination period is half, that is, one second. Further, the values of the other speaker determination parameters shown in FIG. 2 are also used after being multiplied by 0.5.

Example 13. In Example 10 above,
Although the case of inserting the fixed time has been described, the first half of the first half of the next speaker determination period may be deleted. If the same terminal is continuously displayed, video switching may occur earlier, so by shortening the speaker determination period, it is possible to immediately detect a new speaker. Become. That is, in the worst case, the next speaker is not displayed until the second talk is waited for. However, according to this embodiment, by shortening the talker determination period, for example, the next talk is made within 1 second at worst. It becomes possible to display the person. Even when the next speaker is displayed within one second, the speaker displayed up to now is continuously displayed from before that, so that the image switching frequently occurs and it is difficult to see. Absent.

Example 14 In Example 11 above,
The case has been described where the speaker determination default parameter has the same configuration as the speaker determination parameter shown in FIG. 2, but the speaker determination default parameter does not necessarily have to have the same configuration as the speaker determination parameter shown in FIG. However, it does not matter even if there are only some parameters that are to be changed. For example, when it is desired to change the value of the valid voice counter 5c, some default parameters for substituting the valid voice counter may be prepared. Increasing the value of the valid voice counter increases the number of terminals that are determined to be silent. On the contrary, if the value of the valid voice counter is decreased, the number of terminals that are determined to have a voice increases. Alternatively, in the embodiment described above,
Although it has been described that the threshold value may or may not be used, if the threshold value of the zone voice counter 5d and the speaker identification counter 5e is set to 0, no threshold value is used. The same operation is performed, and a parameter for setting the threshold value to 0 may be set as a default parameter, and these values may be rewritten depending on the presence / absence of video switching.

Example 15. In the above embodiment, the case of video switching of the video conference terminal has been described, but the speaker determination in the present invention is not only used for the video conference terminal, but voice signals are input from a plurality of terminals and the input voice signals are input. It is possible to adapt the case of determining the speaker from. In the above embodiment, the case where only one speaker terminal is determined has been described. For example, when two or four video images can be displayed, two or four speaker terminals are selected. But it doesn't matter.

[0112]

As described above, according to the present invention, it is possible to eliminate the inconvenience that a speaker is frequently switched when a voice signal is input from a plurality of terminals and a speaker is judged from the voice signals. Therefore, for example, when the video is switched using the determined speaker, the speaker determination method in which the video is naturally switched can be obtained.

[Brief description of drawings]

FIG. 1 is a block diagram showing a speaker determination method according to an embodiment of the present invention.

FIG. 2 is a diagram showing a configuration example of a speaker determination parameter and a voice count memory.

FIG. 3 is a flowchart showing a flow of operation of the embodiment of the present invention.

FIG. 4 is a diagram showing values of a total voice counter according to an embodiment of the present invention.

FIG. 5 is a block diagram showing an embodiment of the present invention.

FIG. 6 is a diagram showing a concept of zones showing an embodiment of the present invention.

FIG. 7 is a flow chart diagram showing the flow of operation of an embodiment of the present invention.

FIG. 8 is a diagram showing values of a total voice counter and a voice counter for each zone according to an embodiment of the present invention.

FIG. 9 is a block diagram showing a speaker determination method according to an embodiment of the present invention.

FIG. 10 is a block diagram showing a speaker determination method according to an embodiment of the present invention.

FIG. 11 is a block diagram showing a speaker determination method according to an embodiment of the present invention.

FIG. 12 is a block diagram showing a speaker determination method according to an embodiment of the present invention.

FIG. 13 is a block diagram showing an embodiment of the present invention.

FIG. 14 is a flowchart showing a flow of operations of the embodiment of the present invention.

FIG. 15 is a diagram showing an operation when a fixed time is inserted according to an embodiment of the present invention.

FIG. 16 is a block diagram showing an embodiment of the present invention.

FIG. 17 is a flowchart showing the flow of operation of the embodiment of the present invention.

FIG. 18 is a diagram showing a system configuration of a video conference system.

FIG. 19 is a block diagram of a multipoint communication control device.

FIG. 20 is a configuration diagram of a voice addition / distribution unit.

FIG. 21 is a block diagram showing a conventional speaker determination method.

FIG. 22 is a configuration diagram of a speaker determination parameter and a voice count memory in a conventional speaker determination method.

FIG. 23 is a diagram showing a speaker determination period and a voice detection time in a conventional speaker determination method.

[Explanation of symbols]

 1 Speaker determination unit 4 Control processing unit 5 Speaker determination parameter 6 Voice counting unit 7 Voice count memory 8 Effective voice determination unit 10 Higher 2 voice count determination unit 11 Previous speaker memory memory 12 Subordinate zone voice determination unit 13 Video switching Information memory 14 Speaker fixed determination unit 15 Speaker determination parameter change processing unit 16 Speaker determination default parameter

Claims (13)

[Claims] 1. A speaker determination in which voice signals from a plurality of terminals are input, the presence or absence of voice is detected in each voice detection period of a predetermined length based on the input voice signals, and a plurality of voice detection periods are collected. In the speaker determination method that determines the speaker terminal that becomes the speaker for each period, in the previous speaker determination period, the previous speaker storage unit that stores the terminal determined to be the speaker as the previous terminal, and the current speaker In the speaker determination period, a plurality of candidate terminals that are candidates for speaker terminals are selected by comparing predetermined parameters according to predetermined criteria, and candidate terminals selected by the previous terminal stored by the previous speaker storage means are selected. And a speaker determining unit that uses the previous terminal as a speaker terminal of this time, the speaker determining method. 2. The speaker determination means is obtained from the number of voice detection periods determined to be voice among a plurality of voice detection periods in the speaker determination period as a predetermined criterion for selecting a candidate terminal. The speaker determination method according to claim 1, wherein the voice frequency is used. 3. The speaker determination unit is a speaker determination period of a voice detection period determined to be voice among a plurality of voice detection periods in the speaker determination period as a predetermined criterion for selecting a candidate terminal. The speaker determination method according to claim 1, wherein a voiced position obtained from the position inside is used. 4. The speaker determination means is obtained from the number of voice detection periods determined to be voice among a plurality of voice detection periods in the speaker determination period as a predetermined criterion for selecting a candidate terminal. It is characterized by using both the voice frequency and the voiced position obtained from the position within the speaker determination period of the voice detection period determined to be voice among a plurality of voice detection periods in the speaker determination period. The speaker determination method according to claim 1. 5. The speaker determining means compares the previous terminal with the speaker terminal when the difference between the parameters of the previous terminal and the candidate terminal based on the predetermined criterion is less than a predetermined threshold value. The speaker determination method according to claim 1, 2, 3, or 4, wherein: 6. The speaker determination method according to claim 1, 3, 4, or 5, wherein the speaker determination period is divided into zones, and the speaker terminal is determined based on a predetermined standard in units of zones. Speaker determination method. 7. A speaker determination in which voice signals from a plurality of terminals are input, the presence or absence of voice is detected for each voice detection period of a predetermined length based on the input voice signals, and a plurality of voice detection periods are collected. In the speaker determination method that determines the speaker terminal that becomes the speaker for each period, the position within the speaker determination period of the voice detection period determined to be voice among the multiple voice detection periods in the speaker determination period. A speaker determination method characterized in that the speaker terminal is determined based on. 8. A speaker determination in which voice signals from a plurality of terminals are input, the presence or absence of voice is detected for each voice detection period of a predetermined length based on the input voice signals, and a plurality of voice detection periods are collected. In the speaker determination method that determines the speaker terminal that becomes the speaker for each period, the terminal determined as the speaker in the previous speaker determination period and the terminal determined as the speaker in the current speaker determination period A speaker determination method characterized by comprising interval changing means for changing an interval for judging a speaker according to the difference. 9. The interval changing means, when a terminal determined as a speaker in a previous speaker determination period and a terminal determined as a speaker in a current speaker determination period are different from each other,
9. The speaker determination system according to claim 8, further comprising speaker fixing means for delaying the start time of the next speaker determination period by a predetermined time. 10. The speaker determination method includes a speaker determination period storage unit that stores the speaker determination period, and the interval changing unit, the speaker determination period stored in the speaker determination period storage unit. 9. The speaker determination system according to claim 8, further comprising a speaker determination period changing unit for changing the speaker determination period. 11. The speaker determination method according to claim 8, wherein the interval changing means sets a period obtained by multiplying the speaker determination period by a predetermined multiplication factor as a next speaker determination period. . 12. The speaker determination period when the terminal determined to be the speaker in the previous speaker determination period and the terminal determined to be the speaker in the current speaker determination period are the same. 9. The speaker determination method according to claim 8, wherein the speaker determination period obtained by deleting a part of the above is set as the next speaker determination period. 13. A voice signal from a plurality of terminals is input, the presence or absence of voice is detected for each voice detection period of a predetermined length based on the input voice signals, and a speaker determination in which a plurality of voice detection periods are collected is made. In the speaker determination method that determines the speaker terminal that becomes the speaker for each period, the speaker determination parameter storage unit that stores various parameters used for speaker determination, and the value of the parameter stored in the speaker determination parameter storage unit The default parameter storage unit that stores parameters having different parameter values, and the difference between the terminal determined as the speaker in the previous speaker determination period and the terminal determined as the speaker in the current speaker determination period To change the value of the parameter stored in the speaker determination parameter storage unit according to the value of the parameter stored in the default parameter storage unit. Speaker determination method characterized by comprising the meter change means.