JPH024095A - Speaker deciding system for inter-multispot video conference - Google Patents

Abstract

PURPOSE:To switch the picture of a talker through a simple circuit by inputting a pulse to a front protection circuit when the presence of a voice is detected from the pronunciation of the talker, and inputting it to a back protection circuit when the lack of the voice is detected, and controlling an FF by the outputs of both the protection circuits. CONSTITUTION:The voice input 1 of the talker is decided 2 about presence of voice/lack of voice at every prescribed period. A voiced decision output 3 outputs the pulse '1' to an N1-ary counter 5 as the front protection circuit, and unvoiced decision output 4 outputs the pulse '1' to an N2-ary counter 6 as the back protection circuit. During speech, the voiced decision output 3 continues, and when it continues by N1 pieces, it is counted over, and it is inputted to the set input of the FF 13, and outputs a talker decision output 14. When it comes to an unvoiced state after somebody became the talker once, and the pulse '1' continues by N2 pieces, the counter 6 counts over, and resets the FF 13 and outputs a non-talker decision output 15. Thus, prescribed time later after the presence of voice/lack of voice is detected, the talker/ non-talker is decided.

Description

[Detailed Description of the Invention] (Technical Field to Which the Invention Pertains) The present invention relates to a method for determining a speaker from among multiple points in a multi-point video conference by automatically switching to a speaker-to-ground system by detecting audio levels. be.

(Prior Art) Conventionally, speaker determination in this type of multipoint video conference system involves installing a voice detector for each voice input at each point, and using the voice detector to detect whether there is a sound or no sound. When a voice presence is detected, a hangover time corresponding to the duration of the voice presence is added to the voice presence determination and output as voice presence information, and voice presence information about this hangover time is provided for each voice input. This was input into a speaker discriminator, and the same discriminator determined whether the speaker was a speaker or a non-speaker based on the speaker discrimination threshold time based on the voice presence information associated with this hangover time at certain intervals. .

In this method, since speaker/non-speaker determination is performed at regular time intervals, the speaker/non-speaker determination start time and the utterance determination start time are asynchronous, resulting in a drawback that the time required for speaker determination varies. In addition, the time parameters to be controlled include the duration of voice, hangover time, speaker determination threshold time, fixed time speaker/non-speaker determination interval, and the number of control parameters (4), which makes the circuit difficult to control. It had the disadvantage of being complicated.

(Objective of the Invention) It is an object of the present invention to eliminate the above-mentioned drawbacks of the conventional technique and to simplify the control circuit and facilitate control by reducing the number of parameters to be controlled by half compared to the conventional technique.

(Structure of the Invention) (Characteristics of the Invention and Differences from the Prior Art) The present invention includes a voice/silence determiner, front and rear protection circuits, and a set-reset type flip-flop for speaker identification screen switching control. The present invention is characterized in that the speaker/non-speaker determination is performed by setting the time required for the speaker/non-speaker determination by the front and rear protection circuits to be a constant value.

In the conventional technology, the speaker/non-speaker is determined at regular intervals, and the time required for determination varies, but the present invention stabilizes the determination by setting the time required for speaker/non-speaker determination to a constant value. different.

(Embodiment) The figure is a configuration block diagram of an embodiment of the method of the present invention, and this figure shows speaker identification screen switching control in a system that connects multiple video conference rooms through video circuits and audio lines and conducts a multipoint video conference. This is a configuration example.

1 is a voice input, 2 is a sound/silence judge, 3 is a sound judgment output, 4 is a silence judgment output, and 5 is N1 as a forward protection circuit.
6 is an N binary counter as a backward protection circuit, 7 is an N0 binary counter output, 8 is an N binary counter output,
9 is a reset input of the N-ary counter 5, 10 is a reset input of the N-2 counter 6, 11 is a set input of the set/reset type flip-flop 13, 12 is a reset input of the same flip-flop, 14 is a speaker determination output, 15 is a non-speaker determination output.

This method does not immediately identify a speaker even when a voice is detected, but determines a speaker when the number of voices exceeds a certain value (the number of forward protection steps), and once it is determined that there is no voice, it immediately determines the speaker It becomes possible to determine that the person is a non-speaker only when the number of silences exceeds a certain value (the number of backward protection steps). This will be explained below.

It is now assumed that the speaker's voice is input to the voice operator 1.

The speaker's voice is processed by the voice/silence determiner 2, which calculates the voice power at regular intervals and determines voice/silence based on the voice threshold level set in the amplitude direction. The II I I+ pulse is output to the sound determination output 3, and if there is no sound, the It I I+ pulse is output to the silence determination output 4, respectively.

During speech, the voice determination output 3 is a series of l/I II pulses, and the silence determination output 4 is "0" continuously during this period. Therefore, there are N1 i in the N L a counter 5.
Only when the t 1 +1 pulse is input, the count is over, and the II 1 +1 pulse is input to the set input terminal 11 of the set/reset type flip-flop 13, and the N binary counter 6 is reset.

On the other hand, the silence judgment output 4 is N2 because Pa0''' is continuous.
The advance counter 6 does not count up, so the reset input terminal 12 of the set-reset type flip-flop 13
, 110 J+ continues, and the speaker judgment output 14 is '1
In other words, it becomes the speaker's output. In this process, if a short voice such as noise is input to the voice input 1, the N□-base counter 5 does not reach the number of counter stages, so the N0-base counter 5 cannot be counted over. Therefore, the set human power 11 cannot be set to "1", and the speaker cannot be determined.

Also, once you are in the speaker state, if there is silence, N
Since 1'113 pulses are continuously input to the binary counter 6, the N binary counter 6 counts up, and when the number reaches N2, the count is over, and the reset input 12 becomes "1" for the first time. The set-reset type flip-flop 13 is reset, and a non-speaker determination output 15 is obtained. Therefore, if the silent period is short, the N2 counter 6 will not count over and it will not be determined that the person is a non-speaker.

(Effects of the Invention) As explained above, according to the present invention, when a speaker appears from a single point based on the output of the speaker/non-speaker determination, when switching screens at multiple points, , change the screen to the corresponding point, and if there are speakers from multiple points, or if there are no speakers from any of the points, the screen of the point that was being displayed up to that point will continue to be displayed as the speaker point screen. By selecting the forward protection time (Nt) to a predetermined value for the speaker screen switching force to be used, short-term noises will not be used as speakers, so erroneous switching due to short noises from other sources will not occur, and backward protection By choosing the time (N) to a predetermined value,
It is possible to realize a circuit that does not determine a break in a speaker's speech as a non-speaker when a point where the speaker is located is selected as the speaker screen. The above-mentioned effects are the same as those obtained with conventional technology,
Although it is the same.

Since the number of control parameters is reduced from the conventional four to two, which is the control parameter for the sound/non-sound determination output, the circuit configuration becomes simpler.

[Brief explanation of the drawing]

The figure is a block diagram showing the configuration of an embodiment of the present invention. 1...Voice input, 2...Sound/silence determiner, 3
... Sound determination output, 4 ... Silence determination output, 5
...N, advance counter (forward protection circuit). 6...N binary counter (backward protection circuit). 7...N1-base counter output, 8...N2-base counter output, 9...N1-base counter reset input,
10... N binary counter reset input, 11.
・・Set/reset type flip-flop set input,
12...Set/reset type flip-flop reset input, 13...Set/reset type flip-flop, 14...Speaker determination output, 15...Non-speaker determination output. Patent applicant Nippon Telegraph and Telephone Corporation

Claims (1)

[Claims] In a system that connects multiple video conference rooms using video lines and audio lines to conduct multipoint video conferences, an audio detector is installed to correspond to the audio input at each location, and a sound detector is installed for each audio input.
After detecting silence, the voice presence information is input to the forward protection circuit of the frame synchronization circuit used in the multiplex converter, and the silence information is also input to the rear protection circuit, and the outputs of these two protection circuits are used to generate a flip-flop. By controlling the number of steps, the speaker is not immediately determined to be a speaker even when a voice is detected, but is determined to be a speaker when the voice presence information exceeds a certain value of the number of forward protection steps, and once the voice is detected. A multipoint-to-point video conference characterized in that even if silence is detected, the person is not immediately determined to be a non-speaker, but is determined to be a non-speaker only when the number of times of silence exceeds a certain number of backward protection stages. Speaker determination method in the system.