A teleconferencing audio bridge

The present invention relates to communication bridges for use in teleconferencing and to methods for connecting multiple teleconference participants.

Teleconferences allow more than two participants, separated in time and space, to communicate via telephone equipment with terminals and / or bridge equipment or services, such as AT & T's Alliance (R) service. Bridge services typically offer a so-called "meet-me" service, which subscribers dial to join a teleconference, or a dial-out service, where an operator calls the participants in a teleconference. Bridge services connect each subscriber point-to-point with a bridge device that interconnects the multiple subscribers into a teleconference. However, despite the numerous social and economic benefits of teleconferencing, their use has been hampered by the lack of features of the bridge devices and the lack of user control over the transmission and reception of audio signals to or from particular participants.

To solve this problem, Horn has disclosed in U.S. Patent No. 5,113,431, issued May 12, 1992, a system that allows a teleconferencing user to dynamically select particular subscribers whom the user can listen to or speak to during a teleconference. Horn's system gives users some control over the structure of the teleconference. The Horn system does not provide a comprehensive solution for controlling the amplitude of audio signals generated by the participants in the teleconference.

US-A-4558180 relates to multi-point digital teleconferencing using special service circuits. In particular, a multipoint bridge supports multipoint audio teleconferencing and multi-point data communication over voice band channels. A multi-point audio teleconferencing algorithm allows for a fully interactive conference where everyone hears everyone. Variants of this algorithm allow for putting any single branch into a transmit-only or receive-only configuration. In the case of audio conferencing, one obtains the answer for each conference participant by subtracting the conference participant's sample from the conference sum so that he can hear all but himself. For data applications, a master station is allowed to broadcast messages to all other remote stations and receive messages from any remote station. However, remote stations can not communicate with each other but only with the master station. For both of these scenarios, noise and echo control functions are achieved using digital processing techniques.

IEEE GLOBAL TELECOMMUNICATIONS CONFERENCE-GLOBECOM 83, vol. 1, 28.11.1983-1.12.1983, SAN DIEGO (US), pages 548-553, LAWRENCE BARANYAI 'DIGITAL MULTIPOINT TELECONFERENCING FOR SPECIAL SERVICE CIRCUITS', describes a distributed control teleconference speech bridge. More specifically, a line amplifier carries the "talk" signal of a user. The user controls the return keys of the intersections connected to the user's line. In this way, the user controls who can hear the user's voice. Similarly, each of the column amplifiers forms the "listening" path of another user. Each user controls the column switches of the intersections connected to the user's column. By controlling the column switches, each user decides the users to whom the user listens. According to one aspect of the present invention, a communication bridge according to claim 1 is provided.

According to another aspect of the present invention, a method according to claim 12 is provided.

The present invention relates to a teleconferencing bridge arrangement in which at least one subscriber (or user) of a teleconference can control at least one characteristic of input signals coupled through the bridge such that the output signals received by that subscriber in a teleconference include a selected combination of the input signals from all subscribers the changed input signals.

In one embodiment of the invention, a communication bridge consists of N input paths and M output paths defining a logical matrix, with inputs from all the participants being added in a conference to form composite output signals which are fed to each subscriber. To avoid feedback and instability, the input signals from a subscriber to the teleconference in the composite output signals supplied to that subscriber are suppressed. Therefore, each participant in the teleconference receives a different sum of input signals. In the architecture of the bridge, each crossing point of the n input paths and m output paths is assigned a gain element (G), so that the composite output signals (O) supplied to a user i can be represented by the equation Oi = IjGij, where I is the Input signal from a subscriber is. In other words, each output of the bridge is a mixture of all inputs, with each input being multiplied by a parameter selected by the user (ie, gain element). Thus, by allowing each subscriber to vary this parameter for the input signals associated with one or more subscribers, a rich set of audio bridge features can be provided. These features include sub-conferencing, whispering, mute, monitoring, complex gain control, and pseudo-stereo.

Advantageously, the audio bridge features of the invention may be used in a multimedia conferencing environment.

Short description of the drawing

Fig. 1 is a block diagram of a communication network including a teleconference communication bridge embodying the invention;

Fig. 2 shows an implementation of a teleconference communication bridge implementing the invention;

Fig. 3 is an exemplary illustration of a matrix of input and output signals;

Fig. 4 shows a digital signal processor (DSP) which processes input signals using the input / output matrix of Fig. 3,

Figures 5, 6, 7 and 8 are configurations of the input / output matrix of Figure 3 in an arrangement for providing various audio bridge features;

Figures 9 and 10 show two exemplary user interfaces by which teleconference subscribers can control amplification elements in the bridge of the invention.

Detailed description

1 shows a block diagram of a communications network which contains a teleconferencing audio bridge implementing the invention. The communication network of Fig. 1 is arranged to simultaneously support a) multi-user multimedia conferences for digital user terminals and b) single-media analog telephone telephony audio teleconferencing. In an exemplary multimedia application, digitized voice signals are mixed by a bridge 108 which also broadcasts digitized video and data signals over the switching system 107 to multimedia terminals 101, 102, and 103. Similarly, audio signals received from the terminals 101-103 and the telephone sets 104, 105 and 106 (hereinafter the terminals 104-106) are mixed for transmission to subscribers of an audio teleconference. The central nerve of the communications network of Fig. 1 is the switching system 107, which may be either a switch, such as a PBX, or a set of associated exchanges (CO switches) in one or more networks. Thus, the bridge 108 can be concurrently connected to several different networks, thereby facilitating cooperation between users of these different networks.

The switching system 107 communicates incoming audio signals from the user terminals 101 to 106 to the bridge 108 and provides audio signals from the bridge 108 to the user terminals 101 to 106. More specifically, the switching system 107 converts analog audio signals received from the telephone sets 104, 105, 106 into digital format and multiplexes those signals for delivery to the bridge 108 via the digital audio signal device 110 received from the multimedia terminals 101 until 103 are received. The bridge 108 is also connected to a local area network (LAN) 112 via a data device 111. In multimedia applications, the LAN 112 may be used in conjunction with the bridge 108 to allow, for example, the multimedia terminals 101, 102, and 103 to display a single document stored on a server of the LAN 111. The multimedia terminals 101 to 103 may be, for example, integrated workstations and digital telephone sets that can receive and transmit voice, data and video information carrying digital signals. In multimedia applications, the bridge 108 is used as an audio, data, and video bridge. Although the remainder of the present description emphasizes the audio features of the bridge 108, it will be understood that the bridge 108 may also be used as a data and video bridge.

Figure 2 shows a teleconference audio bridge 108 embodying the invention. The bridge 108 is comprised of a network interface 202 which receives incoming signals via the device 110, audio signals over a full-duplex serial link 210 to a DSP 203, and control data signals over a bus 204 to a controller 201 steers. Thus, network interface 202 is the entry and exit point for audio communication and control data signals processed in bridge 108. More specifically, the network interface 202 receives incoming signals via the device 110, determines where these signals are to be routed, and optionally performs signal encoding and framing.

At the heart of the network interface 202 is a Time Division Multiple Access (TSI) 206, which is a space / time multiplex arranged to direct audio signals transmitted over the channels in the device 110 to the DSP 203 (and receiving from this). Control data received from the time slot 206 is forwarded to the protocol processor 207, which packages and unpacks the control data over the bus 204 for transmission to the controller 201.

The DSP 203 is a signal processor consisting of a central processing unit (CPU) 209 and a memory 208 which stores software for implementation of the audio bridge function, including the matrix multiplication to be further described below. The memory 208 includes various input and output buffers and a gain matrix buffer, which will be described in more detail below. The functions of the DSP 203 can be implemented using, for example, a digital signal processor DSP32C from AT & T.

The controller 201 is responsible for processing all control signal data received in the bridge 108. Control signal data may be received in band via the device 110 through a bearer channel (B) of an ISDN Basic Rate Interface (BRT) or via the data device 111 connecting the bridge 108 to the local area network (LAN) 112. As part of its control signal processing functions, the controller 201 accepts signaling messages from users and sets the gain elements for the matrix multiplication function performed in the DSP 203. The controller 201 includes a memory 205 that stores conference control software for calculating the value of the gain elements to implement a specific conferencing function. After a gain matrix has been defined, the controller 201 loads this gain matrix (to be described in more detail below) into the gain matrix buffer of the memory 208 in the DSP 203. FIG. 3 is an exemplary illustration of a matrix of input and output signals.

The matrix of Fig. 3 shows n audio input signals (e.g., speech) Ij (0 ≤ j ≤ n) arriving on the "speaking" columns, and m speech output signals Oi (0 ≤ i ≤ m) appearing on leave the "listening" lines. At each crossing point of the matrix is ​​a reinforcing element Gij. The output signal on a listening line (or bridge function) can be expressed as Oi = IjGIj. In other words, the composite output signal provided to each subscriber Oi of the teleconference is a sum of all voice input signals Ij received from the bridge, each voice input signal being modified by any gain element (Gij) specified by a user or originator of the conference. Another way of describing this bridge function is the matrix-vector multiplication [O] = [I] * [G]. By the participant or the subscribers (or an originator) of the teleconference are allowed to select the values ​​of the gain element (s) associated with one or more particular subscribers, thus the amplitude of the audio signals received from one or more subscribers as desired of this author / participant. For example, the gain elements for G & sub1; & sub1 ;, G & sub2; & sub2 ;, G & sub3; & sub3; set to zero to prevent feedback of a participant's own audio signal. This ensures stability of the composite output signals received by the subscribers 1, 2 and 3.

After deriving output signals in the DSP 203, they are forwarded via the serial link 210 to the timing switch 206, which demultiplexes the composite outgoing signals for delivery to different subscribers. In general, values ​​of the gain elements (G) are limited to (0 ≦ G ≦ 1) in order to restrict truncation of the output signals. The gain elements (G) may take values ​​greater than one to amplify weak input signals.

The bridge of Figure 2 may be implemented in analog or digital form. For example, an analog bridge could be implemented by modifying the analog bridge described in the aforementioned Horn patent.

Horn's analogue bridge uses a) the virtual ground summing node of an operational amplifier to implement each row, and b) CMOS analog switches to implement what may be considered the equivalent of gain elements that are either one (1) or zero (0) are limited. Modification of this bridge would involve replacing the CMOS analog switches (which implement the gain elements) with various analog switches that may take discrete values ​​instead of just the binary values ​​of only one or zero. For example, multiplying digital-to-analog converters could also be used to influence the value of the gain elements.

As is well known to those skilled in the art, a digital-to-analog converter provides an analog output signal (O) equal to a digital input signal (I) multiplied by an analog reference voltage Vref, ie, O = I * Vref. By using the bridge's audio input signal to drive the voltage reference input of the digital-to-analog converter and passing the output signal to a summing node to drive the output of the bridge, a digital signal applied to the digital-to-analog converter can be used to control the gain element in the bridge become.

In an embodiment useful for digital audio signals, the matrix of FIG. 3 is a virtual matrix. The virtual matrix does not physically exist, but is actually represented by a gain matrix stored in memory 208, where each data element represents a crosspoint gain element Gib.

4 shows an implementation of the DSP 203 of FIG. 3. In one implementation of the principles of the invention, digitized input speech signals reach the DSP 203 over the serial link 210. For example, the digitized speech for each input signal consists of a plurality of 125 microsecond samples (one sampling period at the sampling rate of 8 KHz). The input signal samples are represented as an Mu-law compressed 8-bit digital signal word. An input frame contains 32 digital signal words representing 32 input signals. Frames are pushed into and out of the DSP 203 using direct memory access (DMA) control using a ping-pong buffer scheme. Detailed information regarding the ping-pong buffer scheme can be found in the AT & T WE DSP32C Digital Processor Information Manual published January 1990 by the AT & T Document Management Organization.

During each frame period, the incoming frame is temporarily stored in an input buffer 408 and then converted to a 32-bit floating-point format and transferred into buffer 406 (via line 407) for temporary storage. Thereafter, the CPU 209 performs matrix multiplication between the first 32 words of the input frame and an exemplary gain matrix 32 of 32 previously loaded into the gain matrix buffer 402 by the controller 201.

The resulting 32 by 1 output matrix (which is an output frame) is then converted back into Mu-Law Pulse Code Modulation (PCM) signals from the gain matrix 402 via line 410 to buffer 411 and later through the lines one at a time 412 and 413, respectively, into the output buffers 401 and 414. Thereafter, the PCM signals are sent to the serial link 210. During the next frame, the same operation takes place on the alternate set of buffers, ie the input buffer 403 and the output buffer 401, while at the same time the serial input / output by DMA takes place.

Advantageously, this arrangement makes it possible to completely delegate the conference control and matrix manipulation to the controller 201 (which receives control signals through user interfaces described below) while the DSP 203 performs the CPU intensive functions of data shifting, translation and matrix multiplication.

The basic bridge of Figure 3 can be used to implement several practical audio bridge services.

Fig. 5 shows an implementation of the basic bridge of Fig. 3, wherein two independent conferences are executed with a single bridge.

In the matrix of Fig. 5, users 1, 2 and 3 attend a conference while users 4, 5 and 6 participate in a separate and various other conference. Although the example matrix of FIG. 5 shows only two concurrently and independently performed conferences, it is understood that a single matrix can implement any number of independent conferences as long as it is considered that the number of ports does not exceed the size of the matrix.

FIG. 6 shows the example matrix of FIG. 3 where a sub-conference is held as a subset of a larger conference. In the matrix of Fig. 6, the subscribers 3, 4 and 5 can talk privately with each other while they can continue to hear the other subscribers (1, 2, 6) of the reduced level conference. Note that the other participants (1, 2, 6) can not hear the conversation between 3, 4 and 5.

FIG. 7 shows another configuration of the matrix of FIG. 3 in which a subscriber may privately chat with one or more other participants without disturbing the other participants in the conference. This feature will hereafter be referred to as the "Whisper" feature, as it allows a subscriber to whisper something private to one or more subscribers. The matrix of Fig. 7 shows a six party conference in which the user whispers 6 to 1 and 2. This is like an asymmetric subconference. The "whispered" subscribers may choose to privately answer, if they so wish, by dynamically setting the gain elements for their input speech directed to excluded participants to zero.

Fig. 8 shows a configuration of the matrix of Fig. 3, in which each subscriber receives one input (one column) on the matrix and two outputs (rows), ie, left and right. For each output pair, different input signals are mixed with a different ratio in the left and right channels, thereby providing a different position in the apparent stereo sound space. This configuration allows a simple implementation of the pseudo-stereo feature. The latter gives the impression that the voice of each participant in a conference comes from a different point in space. The implementation of the pseudo-stereo feature requires a single input audio channel from each user to the bridge (ie, no stereo microphone is required) and two output audio channels back to the users requesting the feature. Participants listen to the conference via stereo headphones or two speakers. Empirical studies have shown that "pseudo-stereo" techniques improve both speaker-to-speaker intelligibility and speaker identification.

Fig. 8 shows a conference with four participants between A, B, C and D. Each participant hears the other three distributed in alphabetical order from left to right. Note that the value of 0.7 equals half power so that each subscriber appears with the same total power.

Prior to describing the user interfaces for the bridge of FIG. 2, a discussion of a method of ensuring confidentiality is appropriate for each conference participant.

To illustrate the confidentiality problem, consider the scenario in which user A has established a whisper channel to user B. A confidentiality policy is needed to prevent user C from altering the gain matrix to listen to A.

In another example, assume that user N has quietened the gain of user M, either because M appears too loud, or because NM does not want to hear. A confidentiality policy is needed to prevent M (or any other subscriber) from changing the request location of N. The following sets out the rules of a "confidentiality policy" and a mechanism for enforcing those policies.

A first rule allows each user to specify the participant (s) with whom he or she wishes to speak and whom he or she wishes to listen to. For example, an audio path from A to B is established only when A speaks to B and wants to listen to BA. In other words, no one can talk to anyone or listen to anyone if he or she does not want it.

A second rule prevents a participant from interfering with the conversation between two other participants.

To implement these rules, the gain matrix [G] is decomposed into two matrices, the speech matrix [T] and the listener matrix [L], so that Gij = Tij * Lij holds. The [T] matrix is ​​binary with element values ​​of one or zero. The [L] matrix has element values ​​of 0 ≤ L ≤ 1. Each participant may indicate the values ​​of his / her own column on the speech matrix and the values ​​of his / her own row on the listening matrix. That is, participants can specify who they are talking to and at what level they want to hear each other.

Figures 9 and 10 show two user interfaces that allow a user to control the reinforcing elements in the bridge of the invention. The user interface of Fig. 9 allows control of the gain elements for signals received by the bridge from a single console. For example, the user interface of FIG. 9 may be used for conference control in a control mode, ie the conference is controlled centrally by an author / mediator.

The upper left corner of Figure 9 shows a set of call control commands, namely, "Call," "Disconnect," "Refresh," and "Cancel." These commands allow the originator (s) to: a) initiate and terminate connections; and b) associate connections, matrix rows and columns, and user names. The lines that lie on the left side of the matrix indicate the status of the connection for each subscriber, the name and the telephone number of that subscriber. The matrix part of Fig. 9 allows the originator or operator to set each enhancement element, for example, by clicking on the desired matrix entry with a mouse of a workstation. Note that the matrix entries shown in Fig. 9 represent only corresponding gain elements and are not equal to these respective gain elements. Values ​​for matrix entries are between 0 and 1 and become larger / smaller in increments of 0.1. A value of 1 is an attenuation of 0 db (decibels). Each gain decrease of 0.1 under i represents an attenuation of 3 db, so that a gain value of 0.1 equals an attenuation of 27 db. A value of 0 represents infinite attenuation. In an example application of the user interface of FIG. 9 For example, clicking with the left mouse button to enlarge a reinforcing element by one step may be used while the right mouse button decreases a reinforcing element by one step. In this application, the middle mouse button is used to toggle the value of the gain elements between 1 and 0.

The user interface of Fig. 10 shows a user interface when the audio bridge of the invention is used in a multimedia environment. The user interface of Fig. 10 also assumes that the conference is being performed in a peer-to-peer mode, ie, each party may control the gain elements associated with the output signal it receives from the bridge. Each user's display shows a still or moving image of the other participants in the conference. Under the image of each participant is a "Talk" button 1001 and a slider 1002. A user can toggle the PTT button 1001 between on and off to release the transmission of his voice to the person whose image is above the PTT button 1001 to lock. A participant uses the slider 1002 under a person image to adjust the incoming audio volume from that person up or down. Advantageously, the user interfaces of Figures 9 and 10 are simple, intuitive, and fully implement the speech and listening matrices of the gain matrix described above.

Although certain aspects of the present invention have been described only as a linearly controlled system, the amplitude or gain applied to the input signals from a subscriber may also be non-linearly controlled such that the dynamic wave is more consistent from all input signals. In addition, the principles of the invention could also be used to alter the frequency response of the input signals, so that subscribers listening to selective frequencies can be better taken into account.

Although the present invention has been described as applied to an analog signal received over separate audio lines, it could also be used in multiplexed analog or digital systems. Therefore, using the principles of the present invention, input signals received in either frequency multiplexed or time division multiplexed formats can also be altered using either an electrical or an optical signal format.

The above description is to be construed as merely an embodiment of the present invention. Those skilled in the art can readily devise alternative arrangements whose functionality is similar to the present embodiment without departing from the scope of the present invention.