JP2000295286A - Method and device for buffer control for real time voice communication - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prevent occurrence of a missing packet and an interrupted voice signal independently of fluctuation in a network. SOLUTION: A reproduction control module 3-8 monitors a data storage amount of a reception buffer 3-2. When the data storage amount exceeds a threshold L1, a high-speed reproduction module 3-3 thins buffer output packets to obtain high-speed data thereby decreasing the storage amount of the buffer 3-2 and reducing a delay time. When the data storage amount is less than a threshold L2(<L1), the output from the reception buffer bypasses the high-speed reproduction module 3-3 to obtain a usual reproduction speed.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a system for real-time voice communication on a computer network that provides a connectionless packet delivery service such as the Internet or an intranet, for example, a receiving buffer in an Internet voice broadcast or an Internet telephone system. It is related to the configuration of

[0002]

2. Description of the Related Art Conventionally, when audio data and the like are transmitted / reproduced on the Internet in real time, fluctuations in data transmission over a network have occurred, and thus audio data has been lost or interrupted. FIG. 5 shows a system configuration image of a typical Internet telephone.

[0003] A calling terminal 1-1 makes a call to a calling gateway (GW) 1-2 via a telephone line.
The originating GW 1-2 accepts a call from the originating terminal 1-1 and sends guidance to the originating terminal 1-1 when a connection is established. In response, the transmitting terminal 1-1,
The incoming call ID is transmitted to the transmitting side GW1-2 by a method such as PB (push button dial) input. The originating GW1-2 determines the optimal destination GW1-5 based on the destination ID, and starts a connection to the destination GW1-5 via the Internet 1-3. When the connection is established, the receiving GW 1-5 receives the information such as the receiving ID from the transmitting GW 1-2, and makes a call to the partner terminal 1-6 via the telephone line. Since voice data is transmitted over the Internet 1-3 in the form of IP packets, a transmission delay occurs each time routing is performed in the network, and transmission fluctuations are added according to the load on the network. Therefore, the amount of data stored in the reception buffer 1-4 of the receiving side GW 1-5 fluctuates.

[0004]

FIG. 6 shows the receiving side GW1-.
5 shows an example of a change in the amount of data stored in the No. 5 reception buffer 1-4. The transmission rate of the network is modulated by fluctuations of the network, and as a result, if there is a transfer of data exceeding the buffer capacity, the corresponding packet is discarded in part 2-1 in the figure while the buffer capacity becomes zero. , The conversation is interrupted, and the silent part 2-2 continues.

[0005] In addition, TCP is used for real-time voice communication.
Using a transmission protocol that retransmits the missing data as in
P is often used as a transmission protocol. In particular, in networks such as the Internet that are used by unspecified large numbers and in which traffic fluctuations are likely to occur, it was necessary to increase the signal receiving buffer capacity to absorb the fluctuations. As a result, the real-time performance of voice communication is greatly impaired.

In order to solve the above-mentioned problem, there is a technique of measuring a fluctuation amount of a network and adaptively changing a capacity of a reception buffer in accordance with the fluctuation amount. However, when the fluctuation amount is large, it is inevitable. However, the problem that the transmission delay increases is not solved.

[0007]

According to the present invention, in real-time voice communication, high-fidelity playback of voice is also important, but more importantly, it is important that no loss of information or delay occurs. Thinking and suggesting the following method. To prevent the loss, interruption, or increase in delay of real-time data, increase the buffer capacity sufficiently.If the amount of data stored in the buffer exceeds a certain value, adaptively increase the data playback speed. Let it. When the delay decreases to a certain value, the normal reproduction method is changed again. According to the configuration of the present invention, it is possible to reduce the data transmission delay caused by network fluctuations, which is increased due to data loss or loss prevention.

[0008]

Embodiment 1 of the present invention will be described below. Embodiment 1 is a case where the present invention is applied to an Internet telephone. FIG. 1 shows a module configuration of a gateway receiving unit of the Internet telephone system. The real-time voice data received from the Internet 3-1 is
First, the data is stored in the receiving buffer 3-2, and then from the public telephone line 3-6 to the receiving terminal 3-7 via the high-speed reproduction module 3-3, the audio decoder 3-4, and the D / A conversion telephone line terminating device 3-5. Is connected to The high-speed reproduction module 3-3 can be bypassed depending on the reception condition. That is, the switcher 3-9 is controlled by the reproduction control module 3-8, and the reception buffer 3-2 may be connected to the audio decoder 3-4 without passing through the high-speed reproduction module 3-3.

FIG. 2 shows an algorithm of the reception buffer control, and FIG. 3 shows an example of a change in the amount of stored data assumed. The threshold values L0, L1, L2 are set on the reception buffer 3-2.
Is provided. At the start of the reception, the switcher 3-9 is set so that the high-speed reproduction module 3-3 is not operating and bypassed. The reception buffer 3-2 is read by the switcher 3-9.
The read control is stopped by the reproduction control module 3-8, and the operation start and stop of the high-speed reproduction module 3-3 is controlled by the reproduction control module 3-8.

The reception buffer 3-2 together with the reception start 4-1
The received data is stored in the
Is exceeded (from L <0 to L0 or more) in block 4-2.
It is assumed that data transmission is started to the subsequent stage at -3. If the condition of block 4-2 is not satisfied, the data transmission to the subsequent stage is stopped, and in step 4-4, data accumulation is waited again, and the determination in block 4-2 is performed again.

After starting transmission to the subsequent stage, it is determined whether the amount of audio data stored in the reception buffer 3-2 exceeds L1 due to fluctuations in data transfer on the network.
It is judged by 5. If it exceeds, in block 4-6 the switcher 3-9 is switched to the high-speed reproduction module 3-3 to start the operation of the high-speed reproduction module 3-3. If the high-speed operation has already been performed, the high-speed operation is maintained. If the condition of the block 4-5 is not satisfied, the high-speed reproduction module 3-3 is kept bypassed.
Next, in block 4-7, the buffer storage capacity becomes L
If the value is less than 2 (from L2 or more to less than L2), the high-speed reproduction is stopped in block 4-8, that is, the high-speed reproduction module 3-3 is bypassed, and the normal reproduction is started. If the normal reproduction was originally performed, the process is continued. If the condition of block 4-7 is not satisfied, no change is made. In block 4-9, it is determined whether the data reception has been completed.
Return to 2. That is, the above operations are repeated during communication. If it is determined in block 4-9 that the data reception has ended, the data reception ends in block 4-10. In block 4-3, if data has already been transmitted, the transmission is continued. Blocks 4-2, 4-5,
Each determination in 4-7, 4-9, blocks 4-3, 4-6,
Each process of 4-8 is performed by the reproduction control module 3-3. The audio decoder 3-4 decodes and expands the audio data in the packet supplied directly from the reception buffer 3-2 or the packet supplied via the high-speed reproduction module 3-3, and converts the data into continuous digital audio as a DA conversion line termination device. 3-
Supply to 5.

In the example shown in FIG. 3, data reception starts at time t ₀ , and the buffer storage amount exceeds L 0 at time t ₁ .
The reproduction of data is started at a normal speed from the reception buffer 3-2 and transmitted to the audio decoder 3-4. Thereafter buffer fullness increases, beyond the time t ₂ in L1, the reproduction of data from the receive buffer 3-2 fast playback module 3
-3, the buffer accumulation starts to decrease and the delay also starts to decrease. At time t ₃ , the buffer accumulation amount becomes L
2 or less, the reproduction from the receiving buffer is returned from the high speed to the normal speed, and the buffer accumulation amount starts to increase. As described above, when the buffer storage amount exceeds L1, high-speed reproduction is performed, and there is no possibility that packets are continuously lost. If the switching to the high-speed reproduction is not performed, the buffer accumulation amount increases as shown by the broken line in FIG.

FIG. 4 shows an embodiment of the algorithm of the high-speed reproduction module 3-3. In this embodiment, G729A, G732.1.
In consideration of the fact that the audio data is compressed and decompressed for every 30 msec packet, it is assumed that the operation is performed by performing an operation of thinning out this packet at a constant interval. For example, 6
As shown by -1, if one of the three consecutive packets is discarded, 1.5 times speed reproduction is performed. If the packets are discarded by one packet as indicated by 6-2, double speed reproduction is performed. In this high-speed reproduction state, the reproduced voice is in a state of speaking quickly, but unlike the case where packets are continuously dropped, the contents of the call are not partially lost, and the information is sufficiently transmitted.

With the improvement of the reproduction speed, the data on the reception buffer starts to decrease. If the data amount falls below L2 in block 4-6, the high-speed playback module 3-
3 is bypassed and returns to normal speed reproduction.

[0015]

The following effects can be obtained by using the receiving buffer control method. Prevention of loss of received data due to overflow of the reception buffer.
Since the buffer capacity can be made relatively large, it is possible to reduce breaks in audio caused by network fluctuations. Prevent extreme delays.

[Brief description of the drawings]

FIG. 1 is a diagram showing a functional configuration of a first embodiment of the present invention.

FIG. 2 is a flowchart illustrating a reception buffer control algorithm according to the first embodiment of the present invention.

FIG. 3 is a diagram illustrating an example of a change in the amount of data stored in a reception buffer according to the first embodiment.

FIG. 4 is a diagram showing an embodiment of a high-speed reproduction algorithm.

FIG. 5 is a diagram showing a system configuration example of a conventional Internet telephone.

FIG. 6 is a diagram showing the concept of packet discarding and sound interruption caused by network fluctuation.

Continued on front page F term (reference) 5K030 GA18 HA08 HB01 HC01 JT01 KA03 KA19 MA13 MB15 5K051 AA02 BB01 CC01 CC02 DD09 DD13 GG02 JJ11 5K069 AA01 BA03 CA02 CA03 CB08 DB12 FC02 FC11 5K101 KK02 NN07 NN14 NN21 SS21

Claims (2)

[Claims] In a real-time voice communication system on a computer network, a method of controlling a terminal-side reception buffer for absorbing transmission fluctuation added in the network at the time of data transmission. And the amount of data accumulated in the corresponding reception buffer is equal to the threshold L
When the value changes from less than 1 to L1 or more, the data reproduction speed is increased from the normal reproduction speed by the high-speed reproduction module, the data on the buffer is reduced, and at the same time, the delay is reduced. <L1) From above, L2
A buffer control method for real-time voice communication, wherein the playback speed is changed to a normal playback speed when the playback speed changes to less than the normal playback speed. 2. A real-time voice communication system on a computer network, comprising: a terminal-side reception buffer control device for absorbing transmission fluctuation added in the network during data transmission; A high-speed reproduction module for speeding up transmission, a means for switching between a state in which audio data from the reception buffer is transmitted to the subsequent stage via the high-speed reproduction module, and a state in which the high-speed reproduction module is bypassed and transmitted to the subsequent stage. The amount of data stored in the buffer is monitored, and if the amount of stored data exceeds the threshold L1, the state is switched to the high-speed reproduction module, and the amount of stored data becomes the threshold L2 (<L1).
A reproduction control module for controlling the switching means so as to bypass the high-speed reproduction module when the number decreases further.