JPS6010958A - Digital type conference talking device - Google Patents

Abstract

PURPOSE:To improve the quality of conference talking by providing a hysteresis characteristic to an active speaker of maximum value control system speech type conference talking so as to prevent the generation of pseudo switching. CONSTITUTION:A serial voice signal of a conference participant incoming to a PCM reception highway 1 is fed to a frame active speaker detecting circuit FAS4 together with parallel conversion RR, a voice sample maximum participant is detected and fed to a talker order decision circuit SPL10, talker order holding circuit STR11, active speaker decision circuit DASP12, and an active speaker holding circuit ASP13; the result is compared with a specified talker order SPL0 to decide a talker and the voice of the chairman is transmitted to an active speaker from a chairman hold circuit CHA16, active speaker voice hold circuit HASP14, and a chairman voice hold circuit HCHA15 by using a selecting circuit 3 and the voice of the active speaker is transmitted to the remaining participants.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to a digital conference call device that is used by being connected to a digital electronic exchange.

Conventional Structure and Problems Therein FIG. 1 is a schematic block diagram of a digital conference call device generally referred to as a maximum value control type speech type. The configuration of this conventional example will be explained below with reference to FIG. 1. In Fig. 1, l is the reception PCM highway, and 2 is the reception PCM highway. Serial voice samples of conference participants (hereinafter simply referred to as participants) transmitted by the reception PCM highway 1 are parallelized frame by frame. This is an audio sample holding circuit (hereinafter abbreviated as circuit R11) that converts and transmits the converted audio sample to an audio sample selection circuit (hereinafter abbreviated as SEL circuit) 3. The voice samples of the participants transmitted via the reception POM-way 1 are sent to the R, R circuit 2, and at the same time are sent to the frame active speaker detection circuit (hereinafter abbreviated as FAS circuit) 4. will also be sent. F
The AS circuit 4 detects the participant with the maximum audio sample value within the frame (1), and transmits the detection result to the time frame active speaker detection circuit (hereinafter abbreviated as TFAS circuit) 5. The TFA8 circuit 5 detects the participant whose voice sample is the largest in a certain period of time (hereinafter referred to as time frame), and sends the detection result to the time frame active speaker holding circuit (hereinafter referred to as T
(abbreviated as ASP circuit) 6. TASP circuit 6
Then, the participants detected by the TFAS circuit 5 are held. The SDR circuit 3 selects the participant's voice sample held in the TASP circuit 6 from among the participant voice samples, and sends it to the transmission voice sample holding circuit (hereinafter referred to as SD
(abbreviated as R circuit) 7. In the SDR circuit 7,
A voice sample sending circuit (SEND circuit) 8 holds this as a voice sample of the participant (hereinafter referred to as active speaker) to be sent to the sending PCM way, and sends the above voice sample to the sending PCM way 9. Send to.

The 5END circuit 8 learns which participant is an active speaker based on the signal from the TASP circuit 6, sends out a no-call/2-turn NT in the active speaker's reception time slot, and sends the active speaker's audio zangle to the remaining participants. Therefore, now, each participant A, B...
Suppose that a voice sample of H is transmitted, and that participant F is an active speaker.
The arrangement of the signals transmitted to the M...i 1 way 9 is as shown in FIG. 2(b). As described above, in the conventional example, the participant whose voice samples are the largest in each time frame is unconditionally designated as the active speaker. Therefore, there is a problem in that the active speaker may be switched even in response to ambient noise of participants other than the active speaker or an increase in burst sound level, resulting in so-called pseudo switching.

That is, when pseudo switching occurs, noise in a conference call increases, the S/N ratio deteriorates significantly, and the quality of the call significantly deteriorates. On the other hand, in order to reduce spurious switching, it is sufficient to make the time frame long, but if the time frame is made too long,
Since it is unfavorable in terms of call quality, such as deterioration of call quality when the speaker changes, it is practically impossible to make the call length longer.

In addition, in the case of the above conventional example, since the active speaker
Participants outside can only listen to the audio from the active speaker.
Since silence is sent to the active speaker, there is a drawback that the conversation is completely one-way and the progress of the meeting feels unnatural.

OBJECTS OF THE INVENTION The present invention eliminates the drawbacks of the above-mentioned conventional example, and reduces the occurrence of pseudo switching due to ambient noise of participants and increases in burst voice sample values, and improves call quality in conference calls. It is an object of the present invention to provide an excellent conference call device that can be improved.

Structure of the Invention In order to achieve the above object, the present invention determines the speaker ranking for each participant in descending order of the number of voice samples in a time frame, in order of the maximum number of voice samples. When the speaker rank of a participant who is currently an active speaker falls below a predetermined speaker rank (hereinafter referred to as the "defined speaker rank"), the speaker rank will be the highest in the next time frame. The participant whose speaker rank is above the specified speaker rank is registered as the active speaker, and the active speaker will not be switched even if the speaker rank is not the highest.
It has a hysteresis characteristic in the speaker ranking, and furthermore,
The system is configured such that one of the participants is designated as a chairperson by dialing or the like in advance, and a voice sample of the chairperson is sent to the active speaker, and a voice sample of the active speaker is sent to the remaining participants.

DESCRIPTION OF EMBODIMENTS The configuration of an embodiment of the present invention will be described below with reference to the drawings. In FIG. 3, the same reference numerals as in FIG. 1 indicate the same components as in FIG. 1. And 10
is a speaker ranking determination circuit (hereinafter abbreviated as SPL circuit), which selects the participant with the largest voice sample in one frame on the PCM highway sent from the FAS circuit 4, and selects the participant during the time frame. Count every
The speaking order of participants in a time frame is determined in descending order of count number. 11 is a speaker rank holding circuit (hereinafter abbreviated as 8TR circuit);
The speaker rank of the participant detected by the PL circuit IO is sent together with the holder 1 to the active speaker determination circuit (hereinafter abbreviated as DASP circuit) 12. The DA and SP circuit 12 learns the participants who are currently active speakers from the active speaker holding circuit (hereinafter abbreviated as ASP circuit) 13 that holds the current active speakers, and uses the speaker ranking input from the STR circuit 11. By comparing the current speaker rank of the active speaker and the predetermined speaker rank 5PLO, the speaker rank of the active speaker is determined to be the prescribed speaker rank 5.
When the PLO falls below the PLO, switch to the participant with the highest speaker rank in the next time frame, and set the standard speaker rank 5PL.
If the value is O or more, the active speaker is not switched. The active speaker determined by this DASP circuit 12 is sent to the ASP circuit 13 and held there. A in SEL circuit 3
SP circuit 13 and a chairman holding circuit (hereinafter referred to as OH
(abbreviated as A circuit) 16, respectively - knowing the participant who is the active speaker and the chairperson, selecting the voice samples of the two said participants from the voice samples of the participants input through the R and R circuits 2; The active speaker's audio sample is stored in the active speaker's audio holding circuit (hereinafter referred to as HAS).
The chairperson's voice sample is output to the chairperson voice sample holding circuit (hereinafter abbreviated as HOHA circuit) 15. The HASP circuit 14 and the HCHA circuit 15 hold input audio samples, and
Send the audio sample to the 5END circuit 8. The 5END circuit 8 learns from the ASP circuit 13 and the OHA circuit 16 which participants are the active speaker and the chairperson, respectively, and listens to the voices of the active speakers in the time slots on the receiving PCMs of the participants other than the active speakers.

the chairperson's voice sample to the active speaker timeslot. Therefore, the audio sample transmitted from the receiving PCM is as shown in FIG. If so, the arrangement of the audio samples output to the transmitting PCM channel 9 would be as shown in FIG. 4(b). As a result, the active speaker has a hysteresis characteristic in the speaker ranking, and it is possible to reduce the occurrence of pseudo switching due to ambient noise of the participants, burst-like increases in voice sample values, and the like. As a result, it is possible to improve the quality of conference calls.Furthermore, it is possible to set a chairperson for a conference and give the instructions of the chairperson to the active speaker, eliminating the complete one-way nature of the call and making the conference more enjoyable. This allows the meeting to proceed according to the instructions of the chairperson, significantly increasing the naturalness of the meeting.

According to the above embodiment, the frame active speaker detection circuit detects the participant with the largest audio sample within the frame using the POM code as is, and has the advantage that the circuit configuration is simple.

Effects of the Invention The present invention has the above-described configuration, and provides the following effects.

(a) Since the active speaker has a hysteresis characteristic, the occurrence of pseudo switching can be reduced, and the quality of speech can be significantly improved.

(b) Since it is possible to reduce the occurrence of pseudo-switching, the time frame length can be made shorter than in conventional methods, reducing the probability that the beginning of a word will be lost when the speaker changes; Call quality can be significantly improved.

(C) Since the meeting is structured so that a chairperson can be set within the meeting, there is an advantage that the proceedings can proceed more smoothly.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of a conventional digital conference call device, and Fig. 2 is an explanatory diagram showing the arrangement of voice samples in the time slots of each participant on the transmit POM highway and the receive POM highway in the conventional conference call device. , FIG. 3 is a block diagram of one embodiment of the digital conference call device of the present invention, and FIG. 4 shows the arrangement of voice samples in the time slots of each participant in the transmitting PCM highway and the receiving PCM highway in the same embodiment. FIG. ■・Receiving PCM input 2・Audio sample selection circuit, 4・Frame active speaker detection circuit, 5
・・Time frame active speaker detection circuit, 6
Time frame active speaker holding circuit, 7. Sending audio sample holding circuit, 8. Audio sample sending circuit,
9. Transmission PCM number, 10. Speaker ranking determining circuit, 11. Speaker ranking holding circuit, 12. Active speaker determining circuit, 13. Active speaker holding circuit, 14.
・Active speaker voice sample holding circuit, 15 Chairman voice sample holding circuit, 16 Chairman holding circuit. Name of agent Patent attorney Toshio Nakao and 1 other person] Figure 1fXU Figure 2 (αri (b) Figure 33 Figure 4 (0,)

Claims (1)

[Claims] A frame active speaker detection circuit that compares the voice samples of each frame of participants transmitted through the POM highway and detects the participant giving the maximum voice sample; A speaker ranking determination circuit that counts the number of maximum voice samples for each participant and determines the speaker ranking, and when the current speaker ranking of active speakers falls below a predetermined predetermined ranking. ,
A hysteresis system in the speaker ranking that prevents the active speaker from switching if the participant whose speaker ranking is first in the next time frame is registered as the active speaker, but whose ranking is above the predetermined ranking. An active speaker determination circuit having special characteristics, a chair holding circuit that registers one participant determined in advance by dialing etc. as the chair, and an output of this chair holding circuit that holds the audio sample of the chair. A voice configured to send the voice sample of the chairperson to the active speaker determined by the chairperson voice sample holding circuit and the active speaker determination circuit, and to transmit the voice sample of the active speaker to the remaining participants. A digital conference call device characterized by being equipped with a sample sending circuit.