JPH09307651A - Simple stereo transmission system - Google Patents

Abstract

(57) [Abstract] [Problem] To improve the sense of presence by economically enabling stereo communication in a teleconferencing system or the like in which a large number of people participate and are remote. SOLUTION: On the transmitting side, N microphone outputs are fully added to form a batch transmission signal, and at the same time, which microphone is the only one "in use" is detected and this is used as sub information for the batch transmission signal. It is encoded together and transmitted to the ground. On the receiving side, the collective signal and the sub information are separated, and by driving only one of the M output speaker channels by the sub information, a pseudo stereo feeling is created in the space on the receiving side.
Furthermore, an echo canceller is placed to remove the reverberant component that goes around from the speaker to multiple microphones. At that time, only one speaker is driven, and since only a fixed echo path appears from the speaker as a starting point up to the point where a plurality of microphones are fully added, it is sufficient to provide one adaptive digital filter for each output speaker. Conventional ADFs, learning algorithms, etc. can be used as they are.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an audio stereo (multi-channel) transmission method, and more particularly, to an audio stereo transmission method capable of two-way simultaneous communication, which is required in a video conference system or the like.

[0002]

2. Description of the Related Art In recent years, various technologies have been developed and put into practical use that enable natural face-to-face communication between remote places via a communication transmission line. Known techniques of this kind include, for example, a wideband coded transmission technique for voice signals, and an acoustic echo canceller technique for removing howling echoes that interfere with hands-free conversation using a microphone / speaker. . Recently, stereo communication has been desired from monaural communication as a more realistic audio communication system.

As a conventional technique relating to a two-way stereo communication system including an echo canceller, for example, the technique described in Fujii, Shimada: "Structure of a multi-channel echo canceller" IEICE Technical Research Report CS84-178 and the like is known. ing.

FIG. 2 is a diagram for explaining a conventional bidirectional stereo communication system including an echo canceller. The conventional technique will be described below with reference to FIG.

The prior art shown in FIG. 2 is an example configured to perform stereo (two channels) bidirectional transmission.
Microphone input channels 305, 306 (307, 30)
8), that is, a conference room and the like arranged on each side of the east and west having a transmission channel and two speaker output channels 309, 310 (311, 312), that is, a reception channel, four transmission channels. 313-31
6. The connection is made via the echo cancellers arranged on the east and west sides to enable stereo bidirectional full-duplex communication with echo components removed.

The conventional technique shown in FIG.
Since the number of 304 is 2 × 2 = 4 depending on the number of reception channels and the number of transmission channels, an adaptive digital filter (ADF Adaptive Digital Filtration) that creates echo replicas of the echo paths 301 to 304 in the echo canceller.
In essence, four sets of ADFs (hereinafter simply referred to as ADF) are required. In the illustrated prior art, four A
If the DF is in charge of four types of echo paths and operates so that the principle of superposition is established, it is possible to remove the echo component contained in the two transmission channels, and stereo two-way full duplex communication is possible. Becomes

[0007]

As described above, in bidirectional stereo communication according to the conventional technique, the number of ADFs required to realize the acoustic echo canceller is equal to the number of transmission channels and the number of reception channels. In the above example, 2 × 2 = 4 products were required. As is well known, an ADF that simulates an echo path is a long transversal filter that corresponds to the length of the echo path (impulse response length), for example, that handles speech in the 7 kHz band in a room with a reverberation time of 250 ms. Then, if the sampling rate is 16 kHz, a transversal filter of 250 ms × 16 kHz = 4000 taps is necessary, which requires a considerable amount of hardware.

Therefore, the bidirectional stereo communication system according to the above-mentioned conventional technique has a problem that the hardware amount is four times as large as that of bidirectional monaural communication, and the burden of implementation is heavy. Even from the electrical characteristics of the echo canceller convergence speed, {ADF1 and ADF2}, {ADF3 and A
There is a problem that the convergence performance of DF4} is deteriorated.

An object of the present invention is to solve the above-mentioned problems of the prior art and to enable effective two-way stereo communication using a smaller amount of hardware, that is, a smaller number of ADFs. It is to provide a simple stereo transmission system.

[0010]

According to the present invention, the object is to provide N (N ≧ 2) microphone input channels on the transmitting side in a simple stereo transmission system for performing bidirectional full-duplex communication with stereo sound. In-use detecting means for comparing voices to detect which one is in use, and N microphone inputs are fully added to make a transmitted voice, and the transmitted voice is used as main information. The in-use information detected by the sub-information is used as a sub-information, and means for encoding and transmitting both pieces of information through one transmission path is provided, and at the receiving side, the main information is decoded to obtain a decoded voice, and at the same time, the sub-information is obtained. Means for separating information, and for enabling only one of M (M ≧ 2) speaker output channels by the separated sub-information,
An echo canceller having M adaptive digital filters corresponding to M speaker output channels to create an echo replica of the spatial echo path, and subtracting each output replica of the adaptive digital filter from the microphone input full addition signal. It is achieved by providing.

According to the present invention, by providing the above-mentioned structure, it is possible to make a pseudo stereo effect by one speech path and to cancel the acoustic echo to hold a conference between remote places.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of a simplified stereo transmission system according to the present invention will be described in detail below with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of a simple stereo transmission system according to an embodiment of the present invention. In FIG. 1, 1, 2, 21, 22 are microphones, 3, 4, 23, 2
Reference numeral 4 is an A / D converter, 5, 25, 101 and 201 are adders, 6 and 26 are level detection / comparison circuits, 7 is a speech coding circuit (including side information multiplexing circuit), 8 is a monaural transmission line,
9 is a voice decoding circuit (including side information separation circuit), 10,
11 is a switch, 12 and 13 are D / A converters, 14 and 1
Reference numeral 5 is a speaker.

An embodiment of the present invention shown in FIG. 1 is a configuration example of a simple stereo transmission system implemented with a microphone input channel N = 2 and a speaker output channel M = 2. FIG.
In the drawing, voice input / output devices such as a microphone and a speaker, an acoustic echo canceller, a voice encoder / decoder, etc. are depicted symmetrically on the east side and the west side across the transmission path.

In FIG. 1, the member [A] on the east side is now in use.
The entire operation of the embodiment of the present invention will be described by taking as an example the case where the individual speaks.

The transmitted voice of the speaker [A] is collected by the two microphones 1 and 2 and converted into an electric signal. At this time, the transmission signals M1 (t) and M2 (t) from the microphones 1 and 2 have a relationship of approximately M2 (t) ≈M1 (t-Δt) / α. That is, the speaker [A] and the microphone 1,
The level difference α and the phase difference Δt are generated due to the relative distance relationship with respect to 2 and the directivity of the microphone.

For convenient digital processing, these signals are first converted into digital signals by the AD converters 3 and 4 and added by the adder 5, and the transmission main information M is transmitted.
(t) = M1 (t) + M2 (t). On the other hand, the information M1 (t) and M2 (t) are compared in amplitude level by the level detector 6. In the case of the example described, since the member [A] is the speaker, ‖M1 (t) ‖> ‖M2 (t) ‖, and the level detector 6 indicates that the microphone 1 side is "in use". When it is determined that there is one, 1-bit sub information S (t) is output.

In the above description, ‖‖ is assumed to be a time average calculation of electric power. Of course, when all the eastern members are not speaking, multiple members [A],
In some cases, [B] is simultaneously uttered, but the level detector 6
1 without any problem by properly determined decision algorithm
The bit sub-information S (t) can be determined.

The above-mentioned transmission main information M (t) and sub information S (t) are compression-encoded and multiplexed by the encoder 7 and transmitted to the west side via the monaural transmission line 8 in the west direction. There are various techniques for the compression encoding system / multiplexing system, but the present invention may use any system.

The information transmitted to the west side via the monaural transmission line 8 is first separated and restored by the decoder 9 in the west side by the transmission main information M (t) and the sub information S (t). It The decoded audio signal M (t) is supplied to the two output speaker channels, but the output switches 10 and 11 are controlled by the sub information S (t) so that one of the output channels is "in use". It becomes a state. In the case of the example described with reference to FIG. 1, for example, the switch 11 is turned on and the switch 10 is turned off. The audio signal M (t) reaches the members [C] and [D] via the DA converters 13 and 12 and the output speakers 15 and 14. In the case of the example described, the speaker 15 is driven, and the sound image of the speaker [A] is localized at that position. East member [B]
When the speaker speaks, the signal is output from the output speaker 14, and the sound image is localized at the position of the output speaker 14.

As mentioned above, one embodiment of the present invention is
The spatial arrangement stereo feeling of the transmitting side cannot be transmitted to the receiving side, but the receiving side can separate and localize the sound image. Sound localization seems to be a personal feeling,
Although there may be differences in the image of the image depending on the individual, this is sufficient as a sense of presence in a teleconference or the like. However, in video conferences where images are transmitted simultaneously,
Since it is necessary to match the layout of the person on the image with the layout of the separated sound image, it is necessary to adjust the spatial layout of the output speakers.

Next, the operation of the echo canceller will be described.

The sound wave radiated from the speaker 15 or 14 on the west side propagates through the space to reach the members [C] and [D], and is also picked up by the microphones 21 and 22 and becomes an echo input. In order to simplify the explanation, it is assumed here that the members [C] and [D] do not speak. That is, it is assumed that a double talk state (a plurality of people speak at the same time) with respect to the echo canceller is not considered.

The signals M21 (t) and M22 (t) due to the echo input components collected by the microphones 21 and 22 are respectively A
Digitized by the D converters 24 and 23 and added by the adder 2
Added by 5, M20 (t) = M21 (t) + M22
It is assumed to be the fully added echo signal of (t). ADF10
0 and 200 may create a replica of the echo signal M20 (t). In the case of the example described, it is assumed that only the output speaker 15 side is “in use”, so that {speaker 15 to microphone 22} and {speaker 15 to microphone 21} are effective (active Since the echo signals from these paths are added by the adder 25, they are regarded as one echo path combined in parallel.

Therefore, in the case described above, the ADF 100 creates an echo replica in the same manner as a monaural echo canceller. Then, this replica is subtracted by the subtractor 101 from the input component from the adder 25. Then, in this case, the ADF 200 outputs nothing.

If the speaker changes from [A] to [B] on the transmitting side, the output speaker changes from speaker 15 to 14 on the receiving side, and the echo path also changes from speaker 14 to the microphone. ADF activated in response to this
Also changes from ADF100 to ADF200. That is,
Two ADFs of the same number as the microphone input channel N = 2 are modeled corresponding to N echo paths starting from the output speaker channel and ending at N microphone full addition points, and replicas are always generated. It is one of the ADFs.

A well-known learning identification method can be applied to the technique of updating the ADF coefficient using the residual between the echo signal M20 (t) and its replica. Although the level difference detector 26 also performs the comparison operation on the echo component, it goes without saying that it does not make sense if the echo component itself is well erased.

In the above-described embodiment of the present invention, the number of input microphones and the number of output speakers are both N = M = 2.
In the present invention, the number of transmitting microphones and the number of receiving speakers are N = M = 2.
It is not necessary to limit the number to each, and the number can be increased independently. For example, the N microphone inputs can be narrowed down to one and output to any one of the M output speakers. It is also possible to appropriately add a plurality of microphone inputs to analog and then perform AD conversion. Further, a specific hardware configuration can be efficiently realized by using a digital signal processor DSP, which has made remarkable progress in recent years.

According to the above-described embodiment of the present invention, since the N microphone inputs are narrowed down to one on the transmitting side, substantially monaural transmission is possible, and a small amount of sub information indicating the selected microphone is obtained. Is transmitted, it is possible to drive the only speaker on the receiving side to generate a sense of stereo and enable sound image localization. Also, since there is only one output speaker driven on the receiving side, AD for echo cancellation is provided.
It suffices to provide one F corresponding to the output speaker, and a small amount of hardware, which is half A as compared with the conventional technique shown in FIG.
A stereo echo canceller can be realized simply by using the DF.

Generally, A which constitutes an echo canceller
The amount of hardware of the DF depends on the width of the reverberation space and the signal band to be handled. For example, assuming that a 7 kHz band voice is handled in a room with a reverberation time of 250 ms, the sampling rate is set to 16 kHz. In the case, 250ms × 16kHz =
A 4000 tap transversal filter is required. Therefore, it is significant that the number of such long filters can be reduced by half.

[0031]

As described above, according to the present invention, since the N microphone inputs are narrowed down to one on the transmitting side, substantially monaural transmission is possible, and a small number of sub-channels representing the selected microphone are displayed. By transmitting the information, it is possible to drive the only speaker on the receiving side to generate a sense of stereo and enable sound image localization. Also, since there is only one output speaker driven on the receiving side, AD for echo cancellation is provided.
It suffices to provide one F corresponding to the output speaker, and a small amount of hardware, which is half A as compared with the conventional technique shown in FIG.
A stereo echo canceller can be realized simply by using the DF.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a simplified stereo transmission system according to an embodiment of the present invention.

FIG. 2 is a diagram illustrating a conventional bidirectional stereo communication system including an echo canceller.

[Explanation of symbols]

 1, 2, 21, 22 Microphone 3, 4, 23, 24 A / D converter 5, 25, 101, 201 Adder 6, 26 Level detection comparison circuit 7 Speech coding circuit (including side information multiplexing circuit) 8 Monaural transmission line 9 Speech decoding circuit (including side information separation circuit) 10, 11 Switch 12, 13 D / A converter 14, 15 Speaker

Claims (1)

[Claims] 1. In a simple stereo transmission method in which bidirectional full-duplex communication is used for pseudo stereo audio communication, N (N ≧ 2) microphone input channel audio is compared to the transmitting side, which is in use. In-use detecting means for detecting whether or not the input is in use, and N microphone inputs are fully added to form a transmitted voice, the transmitted voice is used as main information, and the in-use information detected by the in-use detecting means is sub-selected. As information, a means for encoding and transmitting both information by one transmission path is provided, and at the receiving side, the main information is decoded to obtain decoded voice, and at the same time, the sub information is separated and the separated information is separated. Means for validating only one of M (M ≧ 2) speaker output channels by sub information, and M corresponding to M speaker output channels for creating an echo replica of the spatial echo path.
A simple stereo transmission system comprising: an adaptive digital filter, and an echo canceller for subtracting each output replica of the adaptive digital filter from the microphone input full addition signal.