JPH0449757A - Conference speech terminal equipment - Google Patents

Abstract

PURPOSE:To maintain speech quality satisfactorily even when a state migrates from a talker state of one person to the one of another person by providing a spread attaching means to attach the spread feeling of an image between image static of two persons. CONSTITUTION:The spread attaching means is comprised of an adder 9, a delay circuit (DEL) 11, a high-pass filter (HPF) 12, a switching circuit 6b, an adder 10, and a substractor 13. The spread attaching means sets a plural talker state where two talkers exist simultaneously within a constant time after a control signal representing a received first sending origin is completed when the state migrates from the talker state of one person to the one of another person, and attaches the spread feeling of the image between the image static of the two persons. Therefore, it is possible to generate atmosphere as if a natural conference is held where senders participate generally as the image static.

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Field of Application] The present invention is applied to a conference call terminal device disposed in a plurality of call terminals interconnected by a conference call audio line.

In particular, the present invention aims to improve the sense of discomfort caused by changes in sound image movement localization when transitioning from one speaker state to another, especially when identifying a calling source from a plurality of callers at an individual caller terminal. The present invention relates to a conference call terminal device that prevents the above.

[Conventional technology]

When communicating with multiple parties, such as in a telephone conference, or when multiple people gather in two conference rooms and have conversations with two parties facing each other, it is difficult to identify who is speaking on a general handset telephone. It has the disadvantage of being unclear and confusing.

Therefore, there is a way to solve this problem by using headphones for binaural listening. That is, a method has been proposed in which a sound image is localized outside the head using headphones. For example, see the book co-authored by Blawelt, Rokubon, and Goto, published by Kajima Publishing Co., Ltd.
"Spatial Acoustics". This method is based on the impulse response 1-1. and H2, and the total impulse response hL (for the left ear), which is preconvoluted with the inverse characteristics Rt and Ra of the loin impulse response from the receiver to the ear canal.
and (for the right ear) are the inputs of the convolutional binaural receiver. However, the total impulse response of the left and right human forces can be expressed by the convolution integral of h=H*R.

FIG. 5 is an explanatory diagram for explaining the extrahead sound image localization technique. 18 is a listener, 17 is a binaural receiver, 21.22.
23.24 and 25 callers ASB, C, D and E
These are the respective sound image localizations.

In FIG. 5, the sound image localization generation method for talking to 6 stations or 6 callers (7) (ASB, C, D, E, F), including myself, will be described below. Note that sound image localization refers to the position of the place where the voice is uttered.

Sound image localization 2 for callers ASBSC, D and E
1. 22, 23, 24 and 25 are arranged to equally divide the sound image localization in front. Normally, it is said that in video conferences and audio conferences with multiple ground connections, an average of six ground connections is common. Therefore, there are five people on the other end of the call or connection to five destinations. As the number n increases, the speaker identification recognition rate from the sound image position decreases. Here, for ease of discussion, we will use 90 degrees to the left, 45 degrees to the left, 45 degrees to the center, and 4 degrees to the right.
5 degrees and 5 places at 90 degrees to the right.

First, the impulse response characteristics from the desired sound image generation position to the listener's outer ear are measured. FIG. 5 shows a case where the speaker F is the listener 18 and the sound image is localized at the location of the speaker C. That is, the impulse response characteristics are Hct and Hcit. The measurement method used has been to measure the impulse response in an anechoic chamber.
Literature (Hayashi “Sound field reproduction using headphones and its subjective evaluation” Auditory Research Group of the Acoustical Society of Japan H-89-19, 1989)
According to the authors, in order for the outer edge of the head to occur naturally, it is better to use a soundproof room than in an anechoic room, since the walls and ceiling behind and on the same side of the subject do not meet the requirements for sound-absorbing materials that do not have strong reflections. It has been reported that L is superior because it is natural and can approximate the actual sound field. Broadband noise is emitted from a loudspeaker, and probe tube microphones or l/8 are placed at the entrances of the subject's left and right ear canals.
A small microphone (about an inch in size) is used to simultaneously convert these three signals from analog to digital. From this, it is possible to calculate the impulse response using the cross-spectral method.

On the other hand, to obtain the inverse characteristics Rt and RR of the headset, the ear canal sound pressure with respect to the electroacoustic conversion input of the headset is measured. Using broadband noise as electrical power for the headset,
A headset is attached to the external ear of the subject, a hole is made in the π1 pad of the receiver of the headset, a probe tube or a 1/8 inch microphone is inserted, and the sound pressure waveform of the external ear is extracted. The electrical input signal and the electrical signal converted from the ear canal sound pressure are converted from analog to digital to obtain inverse filter characteristics. This method is well known as an inverse filter construction method using least squares error in the time domain.

Therefore, the convolved total impulse response is h π R*
It can be found by

Now, if the received audio signal is P and the sound pressure signal of the receiver's external ear force is Q, it is possible to convert the sound #P that is aligned with the receiver's external ear force Q using the total integral consisting of Q = P * h. becomes.

Through the above digital calculation, the voice heard by the caller from the receiver becomes the same sound waveform as if they were in the real space calculated in advance by 91, so as a human auditory truth response, the caller hears it as if it were in the real space. You can talk to the other party at a distance of about 1.5 meters in front of the [ ], and it feels like a natural conversation environment in the real space. Extra-head sensations are part of that natural sensation. Therefore, the user can comfortably enjoy long-term calls without feeling pressured or feeling of auditory or psychological fatigue that is often associated with headset calls.

On the other hand, in the control information means indicating the transmission source together with the transmission audio signal, [['lTT (International Telegraph and Telephone Advisory Committee) Recommendation G, 722, and the paper (Shimada, Meiki: [-Multi-destination audio conference communication system, “Ground Discrimination Sound Image Generation Method” Journal of the Institute of Electronics, Information and Communication Engineers, vol. J70-8. No. 9.198
There is a transmitting light control information encoding method described in 7th year).
This paper presents a theory of sound localization in indoor space using multiple loudspeakers.

In the case of a conference call system in which the audio signals of each speaker are summed and transmitted over an audio line, and where there are multiple speakers at the same time, the presence or absence of audio signals from multiple speakers is determined by the transmission light control information signal. Even if this could be determined, the signal with -degree voice addition cannot be separated into the voice signals of each speaker. Even if separation is possible, the current technology is to recognize the correspondence between the characteristic parameters of the audio signal and the speaker's audio signal (speaker recognition) and generate a sound image at the pre-specified speaker position. It has not reached the practical level.

Furthermore, with the method of localizing sound images outside the head using headphones, when multiple simultaneous sounds occur, the user will move to the center of multiple sound image localizations, which makes the listener feel uncomfortable at 8°. Furthermore, the paper by Shimada et al. only discusses a control method for multiple vocal sound sources, that is, multiple loudspeakers (speakers) that are separated in advance by transmitting optical control listening signals, and does not simply discuss methods for controlling multiple loudspeakers (speakers) that are separated in advance by transmitting optical control listening signals. Receiver drive for both left and right ears! JJ No proposal has been made from the source signal. Therefore, paying attention to the above-mentioned points, one of the inventors of the present application has already developed a method characterized by giving a sense of sound image expansion between multiple sound image localizations when multiple simultaneous talkers occur. It has been proposed (Shimada "Conference Call Terminal Device" Japanese Patent Application No. 112341/1990, filed on April 27, 1990).

[Problem to be solved by the invention]

However, when there is a sudden transition from one speaker state to another speaker state, the impulse response characteristic is convolved with the speech signal, so when such a state transition occurs, There are drawbacks such as noise being mixed into the listening sound at the moment of switching, unnaturalness occurring in the listening sensation of moving sound image localization, and the possibility of deteriorating call quality.

An object of the present invention is to provide a headset conference call device that can maintain good call quality even when transitioning from one speaker state to another by eliminating this drawback. .

[Failure to solve the problem]

The present invention received from the voice line for conference call? T,
sound image generation means for generating sound images for pjH signals 1 and 2 by two electroacoustic conversion means corresponding to both ears of a person;
A conference call terminal device comprising a sound image localization means for localizing the sound image generated by the sound image generation means according to a control signal received together with the audio signal from the audio line,
The sound image localization means is configured to control the sound image localization means, when the state of one person's speaker changes to the state of another person's speaker, for a certain period of time after the end of the received control signal indicating the first transmission light. The present invention is characterized in that it is set to a multi-speaker state in which speakers are present at the same time, and includes a spread imparting means for imparting a feeling of spread of the sound image between the sound image localization of the two speakers.

[Effect]

When the one-speaker state transitions to another one-speaker state, the spread imparting means controls two speakers within a certain period of time after the received control signal indicating the first transmitting light ends. A multi-speaker state exists at the same time, and a sense of spread of the sound image is imparted between the sound image localization of the two people.

Therefore, even when there is a sudden transition from a single speaker state to another single speaker state, sound image localization can create an atmosphere where the speakers are more naturally gathered together for a meeting. It becomes possible.

〔Example〕

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

FIG. 1 is a block diagram showing one embodiment of the present invention.

According to FIG. 1, in this embodiment, a sound image generating means generates a sound image for an audio signal received from an audio line for a conference call using a binaural receiver 17 as two electroacoustic converting means corresponding to both ears of a person. Digital to analog converter (D/A) as 14.2 low pass filters (LPF)
15, and two amplifiers (8 MP) 16, and a signal separation means as a sound image localization means for localizing the sound image generated by the sound image generation means according to the transmission source control signal 32 received together with the audio signal from the audio line. Circuit 1, switching control circuit 2, linear encoding circuit 3, switching circuit (1) 4
a and 4b, a memory circuit 5, a switching circuit (2) 6a, and adders 7 and 8, in which the present invention is characterized in that the sound image localization means A multi-speaker state in which two speakers exist at the same time within a certain period of time after the received control signal indicating the first speaking source ends when the state transitions from one speaker state to another one-speaker state. As a spreading means that gives a feeling of spreading the sound image between two people's sound image localization,
Adder 9, delay circuit (DEL) 11, high-pass filter (
HPF) 12, a switching circuit 6b, an adder IO, and a subtracter 13.

FIG. 2 is an explanatory diagram of the frame format of the source control signal, and shows the divisional use within (iJKbps) of CCITT Recommendation H, 221. The table also shows an example of the code format of the sound image localization control signal.

The basic service of l5DN (Digital Service Integrated Network) includes two 54kbps service information lines (B).
There are 28+D channels with 16kbps service information control line (D> is 1 line.Using the 54kbps digital 1 link line provided between these subscribers, the speed is divided and control is provided to each caller. There is a method for simultaneously transmitting information signals and audio signals to the listener (i.e. recommended by CCITT) I, 221 (G,
? This method uses the frame format of 22). This method is applied to communication systems that can simultaneously transmit not only voice but also digital image signals and data signals using a 54 kbps digital 1-link line. A 64 kbps channel consists of octets (8 bits are one unit) transmitted at 8 kTLZ, and the 8th bit of each octet is called a service channel, and the service channel is further divided into three parts as shown in Figure 2. . The frame alignment signal (FAS) consists of multiple frames in which a basic frame pattern of 8 bits is generated alternately for each frame of the service channel. Bit Allocation Signals (BAS) define various communication modes. Further, the application channel (^C) has 64 bits per frame, and is applicable to data service communication having a transmission rate of 5.4 khps or less.

Table Example of Code Format As an example of implementing this method, the above-mentioned paper by Shimada et al. uses an example of code format as shown in the table as a source control information signal. In other words, the signal from the transmitter is from the transmitter A,
The source control information signals corresponding to B, C, D, E, and F are rlOUOooJ, r010000ko, and ro, respectively.
oiooo,, “000100J,rooooloJ
It is said that

The above paper states that the transmitting source control information signals are encoded sequences that are orthogonal to each other and are desirable because the logical processing can be simplified even when there are multiple transmitters.

Now, the remote conference communication system simultaneously transmits the voice signal of the transmitter and the control information signal of each transmitter. A node device that connects multiple destinations logically processes the control information signal and transmits the audio signal and the source control information signal to the listener side. On the receiving side, the sound is generated in sound image positions divided by direction in advance in response to the sender control information signal, and a sound image is generated as if conference participants were seated at one table. When there are multiple speakers, the transmitting light control listening includes at least a plurality of logic "1" codes.

Next, before explaining the specific operation of this embodiment, the basic operation of this embodiment will be explained using the conference call transition diagram shown in FIG.

In general, the conference call states are as shown in Figure 3: ■ Single speaker state (when there is only one speaker, this state usually occupies 80 to 90% of all conference call states), ■ Multiple speaker state. There are three states: a speaker state (including chatter and laughter of multiple people), and a total silence state (a state in which all parties are not speaking), and the state transitions indicated by the arrows are added to these states. This is the conference call status.

In the case of a multi-speaker situation, various sound image localization methods can be considered, such as adding a sense of spaciousness to the sound image. This method has already been described above. Furthermore, in a completely silent state, instead of setting the sound pressure input to the receiver of the headphones to 0, processing may be performed by adding the room noise and line noise added as they are from each speaker. Therefore, a problem arises when a single speaker, ie, one speaker, suddenly changes state to another speaker.

The basic idea of the present invention is that when the current source control signal transitions from the first speaker to the second speaker, the point at which the first speaker ends is extended for a certain period of time, and the second This gives a sense of spaciousness to the sound image when it overlaps with the starting point of the speaker (
(transition indicated by the dotted line in Figure 3).

The value of this certain period of time is determined, for example, from the time required for the above-mentioned convolution calculation, the unnaturalness of listening, and the like.

This embodiment is constructed based on the above-mentioned basic idea, and when a voice signal is being transmitted from one caller's terminal, the speaking position is determined for each caller according to the received control signal. Generate different sound images. This method enables remote conference communication by connecting different destinations and connecting with multiple callers.By generating sound images in different positions for each caller, it is easy to recognize the callers and send To create a natural conference as if conference participants were seated at one table even when speakers suddenly change.

Next, the specific operation of this embodiment will be explained with reference to the timing chart shown in FIG.

Here, FIG. 4 shows the operation of switching circuits (1) i6 and (2).

In FIG. 3, audio signals and source control information 31 are simultaneously transmitted in Freno format conforming to the CCITT Recommendations G, 722 or H, 221.

Then, an information bit rate of 56 kbps and a control information rate of 3 kbps are separately communicated on a digital transmission path of 54 kbps. The control signal described above is transmitted using this 8 kbps. A signal separation circuit 1 separates these two signals, and the audio signal is converted into a digital nonia code by a linear encoding circuit 3 so that digital signal processing operations can be performed. On the other hand, based on the transmission light control signal 32, the switching control circuit 2 determines whether the voice just received is from a single speaker or multiple speakers. Next, the switching circuit (1)
4a and 4b select an impulse response for generating the speaker's sound image localization corresponding to the audio signal according to the contents of the command from the switching control circuit 2. This impulse response characteristic information (h) is prepared to generate sound image localization outside the head, and is based on the sound image localization of each pre-measured sound image (sound images of callers A, B, C, D, and E in Figure 5). ) is connected to the memory circuit 5 storing the impulse response characteristic information (h) of the switching circuit (
1) Select in 4a and 4b. In FIG. 1, the switching circuit (1) 14&
and 4b are shown.

In the case of a single speaker (one speaker), switching circuit (2) 6
a and 6b are turned off, and are converted into analog signals by the digital analog τ cog converter 14 via the adder circuits 7.8 and 10 and the subtracter circuit 13, and are converted into analog signals to remove quantization noise. Connected to filter 15. Processing information on the extrahead sound image localization is transmitted to the binaural receivers 17 of the headphone set of the listener 18 via an amplifier 16 that makes the volume optimal for the binaural receivers 17 of the headphone set.

Next, an example will be described in which there is a sudden transition from the solo speaker state of speaker A to the solo speaker state of speaker B.

In Figure 4, each speaker A, B, C, DSE and F
The transmitting light control signal 32 corresponding to rlooo
oo”, rolooo(1+, “001000J, ro
ooloo” and “000010 J, so
The transmitting light control signal 32 is initially "rlooooo",
After that, the transition will be "rolooooo". As mentioned above, the switching circuit (1) 4a and 4 at this time
Since the switch position b is connected to the A terminal, hAi and hAL are selected as impulse responses, and convolution with the input audio signal is performed. However, as shown in FIG. 4, if another speaker, here speaker B, starts speaking after a delay of one frame, a sense of discomfort arises in the calculation delay and the movement of the sound image localization.

In order to prevent this, it is assumed that speaker A continues to be in the speaker state from the time when the message to the speaker ends until the end of a certain period of time, and within this fixed period of time, speaker B starts speaking. If this is the case, it is understood that this is a multi-speaker state in which speaker A and speaker B are speakers at the same time, and calculations are performed to give a sense of spaciousness to the sound image.

That is, the switching circuits (1) 4a and 4b turn on the B terminal in the middle of a certain period of time, and at the same time turn on the switching circuit (1) 4a and 4b.
2) Set 6a right and 6b to 1 on. Then, the switching circuit (2) 6a and 6b is activated at the same time as the end of the certain period of time.
"Off".

In this way, by generating a state in which the speaker is on hold for a certain period of time from the information of each transmitting source extracted from the transmitting light control signal 32, it is possible to prevent unnatural movement of sound image localization and performance.
It is possible to prevent the occurrence of hina-song due to the mid-stop of the process.

〔Effect of the invention〕

As explained above, the present invention is advantageous because even if a single speaker in a conference call state suddenly transitions to another single speaker state, the previous single speaker state is suspended for a certain period of time. The sound image localization has the effect of creating a more natural atmosphere in which callers are meeting in a general gathering, and maintaining good call quality.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of the present invention. FIG. 2 is an explanatory diagram showing the frame format of the transmission light control signal. Figure 3 is a transition diagram of the conference call. FIG. 4 is a timing chart showing the operation. FIG. 5 is an explanatory diagram showing a method of extrahead sound image localization. 1... Signal separation circuit, 2... Switching control circuit, 3...
・Linear encoding circuit, 4a, 4b...switching circuit (1
), 5...memory circuit, 6a, 6b...switching circuit (2)
, 7 to 10...addition circuit, II...delay circuit (DP
:L), 12...High-pass filter (HPF), 13
...Subtraction circuit, 14...Digital-to-analog converter (0/^), 15...Low pass filter (LPF), 1
6... Amplifier (AMP), 17... Binaural receiver, 18... Listener, 21-25... Sound image position, 31
. . . Audio transmitter control signal, 32 . . . Transmit light control signal. Patent Applicant Nippon Telegraph and Telephone Corporation Agent Patent Attorney Nao Ide Takashi B 10 B+D 2 B+D
Sentence example (, 4'-Shotsu Kinuta! Fee map) 33 times

Claims (1)

[Claims] 1. Sound image generating means for generating a sound image for an audio signal received from an audio line for a conference call using two electroacoustic converting means corresponding to both ears of a person; and sound image localization means for localizing a sound image according to a control signal received together with an audio signal from the audio line, wherein the sound image localization means changes from a state of one speaker to a state of another speaker. In the case of a transition, a multi-speaker state in which two speakers exist at the same time is established within a certain period of time after the end of the received control signal indicating the first transmission source. A conference call terminal device characterized in that it includes a spread imparting means for imparting a sense of spread to a sound image.