JPH0449756A - Conference speech device - Google Patents

Abstract

PURPOSE:To perform the static of an image at a reception audible side by processing signals from plural mikes for sound, and specifying the position of the mike in accordance with a sender. CONSTITUTION:The sound pressure levels of the mikes M1-Mn are detected by sound detecting means 3a-3d, and a talker position is detected sequentially starting from a first talker by a talker number detecting means, and those talker numbers are stored. Thence, the cross-correlation functions of audio signals received from the first and n-th mikes are found by a cross-correlation function calculation means. Thence, time difference in which the maximum value of the cross-correlation function is found by a maximum value detecting means. Then, match between found time difference and the talker position stored in the talker number storage means is confirmed, and a sender code signal 18 for the talker number is sent out. Therefore, it is possible to perform the static of the image at the reception audible side by attaching a number on the mike.

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Field of Application] The present invention is applied to a conference call device in which a plurality of callers can have a conversation via a communication line.

In particular, the present invention relates to a conference communication system between multiple speakers using multiple microphone inputs. The present invention relates to a conference call device that detects speaker positions.

[Conventional technology]

Conventionally, there is a stereo transmission method in which a plurality of speakers are distributed over two locations and a sound image is reproduced. For example, see the literature (Minami ``Teleconference audio system using pseudo-stereo audio'')
Institute of Electronics and Communication Engineers, Circuit and System Study Group Material CAS86-
29.1986>. As is well known, the transmitter calculates the correlation between the two microphones, sends out this correlation over a separate line, and the receiver uses the two loudspeakers to create the original sound based on the correlation. It is an attempt to revitalize the place.

However, in this method, when there are simultaneous speakers, the amount of correlation becomes undefined and cannot be determined accurately, resulting in unrealistic sound field reproduction. Moreover, even in the case of a single speaker, sound image localization cannot be performed accurately depending on the position of the microphone for collecting sound, the acoustic conditions on the transmitting side and the listening side, especially the reverberation conditions. In the case of a conference call, each speaker is a point sound source, so when a piano is heard from the right side and a violin is heard from the left side, like an orchestra heard in a conventional music stereo broadcast, the sound image localization is ambiguous. This is because they are essentially different.

The reason for this is, for example, that NHK (Japan Broadcasting Corporation) is currently trying to start a high-definition television, or HD TV, service, but the number of loudspeakers is 5.
Three loudspeakers are installed in the front to prevent the sound image from becoming blurred in the center, but this is also due to the desire to make the sound image localization more accurate. However, as mentioned above, in remote conferences, the location of each speaker is an important source of information.
Since the speaker's position is a point sound source, it goes without saying that more accurate sound image localization is required.

As a method to overcome and solve these drawbacks, for example, the literature (Shimada, Suzuki, "Ground identification sound image generation method for multi-point audio conference communication system", Journal of the Institute of Electronics and Communication Engineers, No. J70-
Volume B, No. 9, 1987).

This method proposes a method in which caller terminals are placed at multiple locations and sound images are generated at different positions for each caller terminal. Even if this method is used, it is possible to extend communication between multiple speakers between opposing conference rooms.An example of its application is to assign a caller's number to each speaker in advance. At the same time, the number of the speaker is transmitted at the same time as the voice signal of the speaker, and since the number of the speaker is preset in each loudspeaker on the listening side, the voice signal is generated to the loudspeaker corresponding to that number. If you do so, you will be able to hear the speaker's voice from where the loudspeaker is installed.

[Problem to be solved by the invention]

However, in a conference call system in which multiple speakers are present in one conference room at the same time and multiple microphones are used to collect sound, only the voice of the speaker is not necessarily collected by the corresponding microphone. The reason for this is that even if a microphone is assigned to each speaker, the speaker's voice may be influenced by other microphones depending on the acoustic conditions in the room or the arrangement of the microphone and the speaker.

Therefore, conventional conference call devices that perform remote conference communications such as J'' in audio conferences and video conferences have the following drawbacks.

(1) Since it is difficult to collect the speaker's voice only into the corresponding microphone, considering the acoustic conditions in the room, a microphone is installed for each speaker, and the human voice is collected by the microphone. Only the number of a particular speaker cannot be determined.

(2) In order to prevent the disadvantage of (1), it is considered to use a super-directional microphone, but it is very expensive and the volume may be affected if the speaker moves while talking. Changes in level or even inability to detect sound pressure.

An object of the present invention is to eliminate the above-mentioned drawbacks, and to process signals for voices from multiple microphones when multiple speakers are simultaneously transmitting or when a single speaker is in one conference room. By identifying the location of the microphone corresponding to the caller and assigning a number to that microphone, it is possible not only to localize the sound image on the listening side, but also to enable only the corresponding microphone to operate during the call system. An object of the present invention is to provide a conference call device that can increase the signal-to-noise ratio.

[Means to solve the problem]

The present invention has a plurality of n microphones arranged so that the installation positions are almost in a line corresponding to each speaker and a certain distance away from each speaker, and A digitizing means for converting and outputting an audio signal into a digital audio signal, and a speaker code signal generating means for generating and outputting a speaker code signal for identifying a speaker corresponding to the digital audio signal output from the digitizing means. In the conference call device, the caller code signal generation means includes a plurality of n voice detection means for detecting the sound pressure level from each microphone, and an output signal output from the plurality of n voice detection means. one stage of speaker number detection memory for detecting the order of speakers and storing the corresponding speaker numbers;
a cross-correlation function calculating means for calculating a cross-correlation function of the audio signals received from the second two microphones; and calculating a time difference at which the inter-eye function is maximum from the cross-correlation function calculated by the cross-correlation function calculating means. maximum value detection means;
It is confirmed that the time difference calculated by the maximum value detection means matches the position of the speaker for the speaker number stored in the speaker number detection storage means, and the speaker code signal for the speaker number is detected. and confirmation means for sending.

Further, in the present invention, the cross-correlation function calculation means can be a means for dividing all n microphones into small sections each consisting of m microphones, and calculating a cross-correlation function for each m section.

[Effect]

In the conference call device of the present invention, the method of detecting the positions of multiple simultaneous speakers is based on the time difference that is the maximum value of the cross-correlation function of the audio signals received from two microphones, and the time difference between the sound source, that is, the speaker and each microphone. This is based on the fact that it corresponds to time differences.

The sound pressure level from each microphone is detected by the voice detection means, and the speaker positions are sequentially detected from the first speaker position by the speaker number detection and storage means, and the speaker numbers are stored. Then, the first and nth
The cross-correlation function of the audio signal received from the second microphone (or the microphones at both ends of each sub-section by dividing n microphones into m sub-sections) is calculated, and the cross-correlation function is determined by the maximum value detection means. A time difference having a maximum value is determined, and a confirmation means confirms that the determined time difference matches the position of the speaker stored in the speaker number storage means, and a speaker code signal for the speaker number is determined. Send.

Therefore, by assigning numbers to the microphones, it becomes possible to localize the sound image on the listening side. Furthermore, by dividing the microphones into m categories, it is possible to operate only the corresponding microphones to increase the signal-to-noise ratio in the communication system.

〔Example〕

Embodiments of the present invention will be described below with reference to the drawings.

First, the basic principle of the present invention will be explained, and then specific embodiments will be explained.

FIG. 1 is an explanatory diagram showing part 9 of the first embodiment of the present invention (/,), showing the relationship between the arrangement of microphones and the seat position of the speaker.

Here, Sl, , Sl % , S, , are speakers, and the corresponding M 1 % , Mi s , L are microphones. Let the distance between the speaker and the microphone be r, and the distance between the microphones be d.

The basic principle of the present invention will be explained below using FIG. Now, assuming a free-diffusion sound field 1, we will make basic considerations regarding the case of a single speaker and the case of multiple speakers.

FIG. 2 shows an example of the cross-correlation function between microphone M1 and microphone Mh.

(1) In the case of a single speaker (single sound source) The sound pressure waveforms Ri(t) and Rn(t) humanly input from the sound source of the speaker SI to the microphones M1 and M, , are generally given by , where A ([11> is the amplitude line segment, φ(ITI
) indicates the phase component.

The cross-correlation function Cc, (t, τ) is is given by

Cross-correlation function C of microphones M1 and M, , corresponding to each speaker
The time difference τ at which c, (t, τ) is maximum is r = TI
It is ITin. That is, in FIG. 1, when 1d=nd/2, in other words, L, -1.

When , then -Tin, r=Q, and a cross-correlation function like the dotted line shown in FIG. 2 is obtained. Therefore, it turns out that the sound source position on the speaker axis is at the center of the positions of microphones M1 and M, and at that time, equation (1) takes the maximum value (in the case where there is one local maximum value).

(2) When there are multiple sound sources (especially in the case of two sound sources), the sound pressure waveforms Rl (t) and R, (t) that are manually input from the sound sources of speakers Sl and S to microphones M and M7 What is "'?
L.

Correlation function Ccr(
t, r) is given by.

Microphone M for each sound source of speaker Sl and speaker S.

The value of the time difference in the first term where the cross-correlation function c ct (t, τ) between and M is the maximum value is τ = T' i 1 T i h The value of the second term is τ - TJI Tj be.

Now, if the speed of sound is C and the distance between the speaker SL and the microphone M1 is, then the delay time from the sound source of the speaker 8 to the microphone M is 1゛1. is given by T I l = L L l / c = EPT〒F/C.

Similarly, when there is a sound source from multiple speakers, as can be seen from equation (2), the time difference between the multiple maximum values, that is, the local maximum value, corresponds to the time difference from the sound source to each microphone. The location will be determined from the time difference. The solid line in FIG. 2 shows an example in which speakers SI and S sit in the center and become speakers at the same time. That is, this is the case of (LiLL・Yu・)−1−・jl Lj・.

The above is an outline of the operation in a free sound field, that is, a room with no reverberation time, but even if there is a reverberation time, reverberation does not affect the initial response part of the signal impulse response, but affects the subsequent part, so the main point is There is no essential influence on the method of the invention. In addition, in the case of diffuse indoor noise, there is no temporal correlation, so equation (2) holds true (7), making it a robust method against noise.

FIG. 3 is a concrete block diagram of the present embodiment. In the present embodiment, the installation positions are arranged almost in a row corresponding to each speaker M, to M, and each speaker M1 to M,
A plurality of n microphones 1a arranged at a certain distance from
1d, and each speaker M+ input from each microphone la to Id.
Analog-to-digital converter 2, delay circuits #14a to 14d, and AND circuits 15a to 15d, which serve as digitizing means for converting and outputting the audio signal of ""-M- to a digital audio signal 17, and from this digitizing means. In a conference call device equipped with a speaker code signal generating means for generating and outputting a speaker code signal 18 for specifying a speaker corresponding to an output digital audio signal 17, the present invention is characterized by the following: The speaker code signal generation means includes a plurality of voice detection circuits 3a to 3d as voice detection means for detecting the sound pressure level from each of the microphones 1a to 1d, and voice detection means 38 to 3d.
A first speaker detection circuit 4 and a number storage circuit 5 as speaker number detection and storage means for detecting the first speaker from an output signal output from the first speaker and storing the corresponding speaker number; First and nth two microphones 1a and 1d
The shift register circuits 6a and 6b, the delay time sweep circuit 7, the variable delay circuits 8a and 3b, and the multiplication circuit 9 as cross-correlation function calculation means for calculating the cross-correlation function of the audio signal received from the An integrating circuit 10 and a maximum value detection circuit 11 as maximum value detection means for determining the time difference at which the cross-correlation function is maximum from the cross-correlation function calculated by the function calculation means, and the time difference calculated by the maximum value detection circuit 11. The correspondence between the speaker number stored in the number storage circuit 5 and the speaker's position is determined *
The system includes a position detection circuit 12 and a matching circuit 13 as confirmation means for transmitting a speaker code signal 18 corresponding to the speaker number.

Next, the operation of the present embodiment will be explained.

First, to simplify the explanation, in FIG. 1, assume that the speakers are SI and SJ, that speaker S1 is the first speaker, and that speaker S4 is the second speaker. Mike M
As is clear from FIG. 1, the audio signal reaches i earlier than the other microphones. As a result, each voice detection circuit 3
For a to 3d, the earliest time to detect the sound pressure level is
The voice detection period in FIG. 3 is ¥83b, and this is detected by the first speaker detection circuit 4 as the first speaker. Here, the microphone number is written in the smallest number in the number storage circuit 5.

Next, when the caller axis becomes the speaker, the sound pressure level is also detected by the voice detection circuit 3C, and this is stored in the number storage circuit 5 as the second speaker. Of course,
At this time, if the speaker Si becomes the listener, the sound pressure level of the voice detection circuit 3b decreases, so the speaker S becomes the first speaker.

The address of the number storage circuit 5 will be written. Therefore, the first speaker always occupies the lowest number in the number memory circuit 5, and when all the speakers become listeners, the microphone number is not written in the number memory circuit 5 and is in a clear state. It becomes. Therefore, if the contents of the number storage circuit 5 are examined, the first and second speakers detected by the voice detection circuits 3a to 3d are always registered in the order of their line numbers.

On the other hand, Mike M1 and Mike M. In order to correlate the audio signals from the input terminal, the audio signals are manually input to shift register circuits 6a and 6b which store audio signals of a certain time interval. The distance difference between the speaker and the microphone showing the maximum delay time difference, that is, the maximum delay time rllaM of the variable delay circuits 8a and 8b in FIG. This is the time calculated by dividing the distance difference by the speed of sound. Here, the variable delay circuits 8a and 8h are interlocked with each other, and when both delay times are 0, τ of the cross-correlation function in equation (2) indicates 0.

When the variable delay function 8a sweeps the delay time, the delay time of the variable delay function 8h remains at 0. Conversely, when the variable delay function 8b sweeps the delay time, the variable delay function 8a The delay time is 0 and it is stationary. Zunai, -
When τ1, 8, the delay time amount of the variable delay circuit 8a is maximum, and when the variable delay amount τ decreases to a value of 0, the delay time @T of the variable delay circuit 8h increases, and the maximum delay is reached. A command is given using the delay time sweep function 7 to sweep up to a time of 10τ□9.

A maximum value detection circuit that multiplies the output audio signals of the two variable delay circuits 8a and 8b in a multiplier circuit 9, passes through an integration circuit 10 that performs integration over a time interval T, monitors the integrated value, and has a comparison function. When the value becomes lower than the previous value in step 11, the microphone position corresponding to the delay time of the delay time sweep circuit 7 at that time is transmitted to the position detection circuit 12 which stores the position, and the position information of the microphone is transmitted to the number storage circuit 5. Output signal and matching circuit 1
Verify with 3.

Furthermore, each microphone M, , 9M, , , Mj, , Mn
The audio signal outputted from the delay circuit groups 14a to 14 has a delay time equal to the processing delay time necessary for audio detection and the processing time necessary to detect the maximum value of the cross-correlation function.
Based on the microphone position information obtained from the output of the matching circuit 13 via d, only the audio signal of the corresponding microphone is passed through the AND circuits 15a to 15d, and further, when there are multiple speakers, addition is performed. They are summed by a circuit 16 to obtain a digital audio signal 17. Further, the signal from the microphone M corresponding to the speaker S is also output as the speaker code signal 18.

By the way, if speaker S is so far away from microphone M that the sound pressure cannot be detected, speaker SI is far away from microphone M1, which is at the end.
The input may not reach a sufficient sound pressure level as an audio signal.

In the second embodiment shown in FIG. 4, the relationship between the microphone position and the speaker position is determined so as to increase the signal-to-noise ratio.

That is, calculation of the cross-correlation function between microphone M and microphone Mb, microphone M and microphone M. By independently calculating the cross-correlation function in the interval between and the cross-correlation function in the interval between microphone Md and microphone M, it is possible to obtain a large signal-to-noise ratio, resulting in high accuracy. It becomes possible to detect the speaker's position.

In the explanation of the specific operational block diagram, the difference from FIG. 3 is the shift register circuits 6a and 6b.

Delay time sweep circuit 7, variable delay circuits 8a and 8h, multiplication circuit 9, integration circuit 10, maximum value detection circuit 11, O, L
The position detection circuit 12 may be configured to operate in a plurality of configurations.

Further, FIG. 5 is an explanatory view showing the main part of a third embodiment of the present invention, and shows the relationship between the microphone position and the speaker position when the present invention is applied to a desk with an arcuate structure used for a video conference. So far, we have explained the rows of microphones in terms of straight lines, but the essential operation is the same even if the rows of microphones are medium-sized.

〔Effect of the invention〕

As explained above, the present invention detects the first speaker who is manually input to a plurality of microphones, and also determines the time difference at which the cross-correlation function of the input audio signals of the two microphones is maximum. Since the position is known, by comparing it with the voice detection means, it is possible to detect the position of the corresponding microphone of the speaker even in a conversation state where there are multiple speakers at the same time. When a person is simultaneously a transmitter or a single speaker, by assigning a name to the microphone, it not only becomes possible to localize the sound image on the listening side, but also activates only the corresponding microphone to improve the communication system. This has the effect of increasing the signal-to-noise ratio.

[Brief explanation of the drawing]

FIG. 1 is an explanatory diagram showing the main parts of a first embodiment of the present invention. FIG. 2 is a characteristic diagram showing an example of the cross-correlation function between the two microphones. FIG. 3 is a detailed block configuration diagram thereof. FIG. 4 is an explanatory diagram showing main parts of a second embodiment of the present invention. FIG. 5 is an explanatory diagram showing main parts of a third embodiment of the present invention. 1a to 1d... Microphone, 2... Analog-digital converter, 3a to 3d... Voice detection circuit, 4... First speaker detection circuit, 5... Number storage circuit, 6a, 6b.
...Shift register circuit, 7...Delay time sweep circuit,
8a. 8b... variable delay circuit, 9... multiplication circuit, 10...
・Integrator circuit, 11... Maximum value detection circuit, 12... Position detection circuit, 13... Matching circuit, 14a to 14d...
・Delay circuit, 15a to 15d...AND circuit, 16・
... Addition circuit, 17... Digital audio signal, 18.
...Speaker code signal, M, -M, ...Microphone, S, ~
Sh...Speaker. Patent Applicant Nippon Telegraph and Telephone Corporation Agent Patent Attorney Nao Ide Correlation function”) Ryo 2 Figure

Claims (1)

[Scope of Claims] 1. A plurality of speakers arranged corresponding to each speaker so that the installation positions are almost in a line and separated from each speaker by a certain distance.
, a digitizing means for converting and outputting the audio signals of each speaker inputted from each microphone into digital audio signals, and a transmitter for identifying the speaker corresponding to the digital audio signal output from the digitizing means. In the conference call device, the caller code signal generation means includes a plurality of n voice detection means for detecting the sound pressure level from each microphone. , speaker number detection storage means for detecting the order of output signals output from the plurality of n voice detection means and storing the corresponding speaker numbers; and two arranged first and nth voice detection means. cross-correlation function calculation means for calculating a cross-correlation function of an audio signal received from the microphone; maximum value detection means for calculating a time difference at which the cross-correlation function is maximum from the cross-correlation function calculated by the cross-correlation function calculation means; It is confirmed that the time difference calculated by the maximum value detection means matches the position of the speaker for the speaker number stored in the speaker number detection storage means, and the speaker code signal for the speaker number is detected. 1. A conference call device comprising: confirmation means for transmitting a message. 2. The conference call device according to claim 1, wherein the cross-correlation function calculation means divides all n microphones into small sections each consisting of m microphones, and calculates the cross-correlation function for each m section.