JPS6365181B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a voice packet transmission delay control system for absorbing variations in the transmission delay time of voice packets in a voice packet switching system.

In recent years, a voice packet switching system has been realized in which voice is digitized using a PCM method or the like, and further converted into packets and sent and received via a packet switching device. Figure 1 is a schematic diagram of a voice packet switching system, where 1 and 2 are telephones, and 3 is an A/C converter that converts analog signals from telephone 1 into digital signals.
D conversion device, 4 is a voice/silence detection device, 5 is a packet assembling device, 71, 72, . 9
is a D/ that converts digital signals to analog signals.
The A converter 10 is a delay time fluctuation absorber, and the telephone 1, A/D converter 3, utterance/silence detector 4, and packet assembly device 5 are on the transmitting side, and the delay time fluctuation absorber 10 and packet Decomposition device 8,
A D/A converter 9 and a telephone 2 represent the receiving side. The voice signal from the telephone 1 is digitized by the A/D converter 3, and further divided into transmission units by the voice/silence detector 4 to determine the presence or absence of voice.
The sound-only packet assembler 5 performs processing such as adding header information (address information, etc.), converts it into a packet, and sends it to the packet switching network 6.
In the packet switching network 6, the number of relaying packet switching devices varies depending on the setting of the call from the sending side to the receiving side, and voice packets are transferred to the packet switching device 71.
~7n is relayed and transmitted to the receiving side. On the receiving side, the audio packets are decomposed by a packet decomposer 8 and further decoded into audio signals by a D/A converter 9.

In a packet switching system, before a voice packet sent from the transmitting side reaches the receiving side, it mainly undergoes packet assembly in the packet assembling device 5 on the transmitting side, and disassembly and transmission in the packet disassembling device 8 on the receiving side. Required fixed delay time and relaying packet switching device 7
1 to 7n, and the latter delay time is not constant. Fluctuations in delay time in voice packet transmission not only cause audio deterioration, but in extreme cases may even make conversation impossible, so it is necessary to absorb fluctuations in delay time when receiving voice packets. This is indispensable, and in the past, variations in delay time have been handled by the delay time variation absorbing device 10.

Conventionally, as a method for absorbing variations in the transmission delay time of voice packets in the delay time variation absorbing device 10, received voice packets are sequentially stored in a buffer memory and the delay time of each voice packet is corrected to a certain fixed value. , a method was used to compensate by changing the delay time for each talk spurt (sound part) of the audio time. In the former method, that is, a method in which the delay time of each received audio packet is corrected to a constant fixed value D, the audio packet input to the delay time fluctuation absorbing device 10 on the receiving side at a transmission delay time D ₁ is stored in the buffer memory ( _Similarly , audio packets received with transmission delay time Do are allowed to stay in the buffer memory for (D- _{D o} ) time, thereby correcting the delay time of each audio packet to a constant value D. This absorbs fluctuations in transmission delay time, and audio packets with a transmission delay time of D or more are discarded because they cause deterioration in audio quality. However, the distribution of delay times of voice packets in a packet switching network becomes a gamma distribution as shown in FIG. 2, assuming that the average transmission delay time per relay packet switching device is equal. In Figure 2, the horizontal axis shows the transmission delay time (t) and the vertical axis shows the cumulative distribution of delay time (%).
3 and 4 correspond to the number of relay packet switching devices, respectively. In the same figure, two curves are shown corresponding to the number of relay packet switching devices, and these curves indicate differences depending on the size of the average transmission delay time for each number of relay packet switching devices. When the average transmission delay time per relay packet switching device is equal, when comparing the same transmission delay time, the cumulative distribution is higher when the number of relay packet switching devices is small than when it is large; It is clear that the variation in transmission delay time is small when the number of packet switching devices is small, and on the contrary, the variation in transmission delay time is large when the number of relay packet switching devices is large. Therefore, this method cannot cope with variations in transmission delay time due to the number of relay packet switching devices, and when the variation in transmission delay time in packet switching network 6 is small, the time required to store voice packets in the buffer memory becomes unnecessarily long. On the other hand, when the transmission delay time fluctuates greatly, the packet loss rate (the probability that an audio packet is not in the buffer memory at the time when it should be taken out) increases.
However, there were drawbacks such as high audio quality and poor audio quality.

In addition, in the latter method, that is, a method in which the delay time is changed for each talk spurt (sound part) of the audio signal to compensate for variations in transmission delay time, the packet switching network 6
This relationship is used assuming that the distribution of transmission delay times in 2 follows the exponential distribution shown in FIG. In other words, as in Figure 2, Figure 3 shows the relationship between the cumulative distribution and the transmission delay time (t) of voice packets, with the horizontal axis as the transmission delay time (t) and the vertical axis as the cumulative distribution (%). Further, the axis is set opposite to the scale of the cumulative distribution, and the packet loss rate is 0 to 100% (packet loss rate is 0% for the cumulative distribution of 100%, packet loss rate is 100% for the cumulative distribution of 0%), and the audio signal is The transmission delay time distribution curve is determined by determining the average transmission delay time for each talk spurt (sound part), and the transmission delay time corresponding to the predetermined packet loss rate is determined (a in Figure 3). →b→c), the transmission delay time is set as the delay time D to be corrected, and the correction is performed using the method explained in the former method when absorbing the variation in the transmission delay time of the voice packet of the next talk spurt. be. This method controls the delay time so that the packet loss rate is constant according to the transmission delay time distribution, so it can obtain better call quality than the former method, but this is explained in Figure 2. As shown above, the distribution of transmission delay time in a packet switching network is a gamma distribution that depends on the number of relay packet switching devices and the average transmission delay time, so when approximated by the exponential distribution shown in Figure 3, the actual delay time The difference in the distribution was large, and the error in determining the delay time was large, making it impossible to keep the packet loss rate constant and making it difficult to maintain call quality.

The purpose of the present invention is to eliminate these drawbacks.
Assuming that the distribution of transmission delay time, which fluctuates in a packet switching network, follows a gamma distribution, the transmission delay time is measured on the receiving side, and the packet loss rate is reduced by controlling the delay time according to the transmission delay time distribution. It is designed to maintain the call quality by keeping it constant.
This will be explained in detail below.

FIG. 4 shows an embodiment of the present invention, which corresponds to the delay time fluctuation absorbing device 10 in FIG.
1 is a relay packet switching device number input terminal, 12 is a relay packet switching device number receiving circuit, 13 is a processing device, 14 is a voice packet input terminal, 15 is a buffer memory for absorbing delay time fluctuations, 16 is a voice packet reception time measuring circuit, 17 is a timer circuit, 18 is a delay time storage section, 19 is a control circuit, 20 is a timer circuit, and 21 is an audio packet output terminal. To operate this, first, when setting up a call, the number of relay packet switching devices is input to the number of relay packet switching devices input terminal 11, and the number of relay packet switching devices receiving circuit 12 receives and processes the number of relay packet switching devices. device 13. In this embodiment, since voice packets are transferred through the same route for one call, the number of relay packet switching devices can be easily known at the receiving station using various systems such as the common line signaling system. Audio packets input from the audio packet input terminal 14 are sequentially input to a buffer memory 15 for absorbing delay time fluctuations and stored therein. The voice packet reception time measuring circuit 16 uses a timer circuit 17 to measure the reception time every time a voice packet is received, and provides it to the processing device 13. Using the reception time of the relay packet, the processing device 13 calculates the average value of the difference between the reception times of two consecutive voice packets in the talk spurt (sound part) of the voice, and further calculates the average value of the difference between the reception time of the relay packet and the relay packet exchange. Depending on the number of devices,
A delay time storage unit 18 consisting of a read-only memory that stores the average value of the difference in reception times and the delay time corresponding to the number of relay packet switching devices shown in FIG.
, and provides a delay time value to the control circuit 19 according to the stored contents. In addition, Figure 5 shows the fourth
This figure shows the stored contents of the delay time storage unit 18, and the number of relay packet switching devices 1, 2, ..., N and the average value E ₁ of the difference in reception time of two consecutive voice packets in the talk spurt, E ₂ ,…,
Delay times D ₁₁ to D _NM corresponding to _EM are stored.
The control circuit 19 checks whether or not audio packets are stored in the buffer memory 15 for absorbing delay time variations, and if not, sets the delay time given by the processing device 13 as a new delay time. The control circuit 19 also uses a timer circuit 20 to store the first received audio packet after the updated settings in a buffer memory for absorbing delay time fluctuations by the updated delay time, and then stores the stored audio packets in a buffer memory for absorbing delay time fluctuations. The audio packets are output to the audio packet output terminal 21 using the first-in-first-out method, and variations in the transmission delay time of individual audio packets are absorbed.

Next, a method for controlling the delay time will be explained in detail. FIG. 6 is an explanatory diagram showing the transmission state of the audio signal,
The high part of the square wave is the talk spurt (sound part)
, the low parts of the square wave indicate pauses (silence), and the upward arrows in each talk spurt each represent a voice packet. In the figure, the receiving side receives the audio packet of the nth talk spurt, calculates the delay time using the method described later, and then
At the time of the pause, it is confirmed that no audio packets are accumulated in the buffer memory 15 for absorbing delay time variation, and the delay time is updated, and the audio packets of the next (n+1)th talk spurt are updated according to the newly updated delay time. fluctuations are absorbed. Also nth
If the th pause is short and the delay time variation absorbing buffer 15 is not empty, the delay time is not updated and the variation in the transmission delay time of the voice packet of the (n+1)th talk spurt is absorbed. Hereinafter, a method will be described in which the processing device 13 determines the delay time distribution in the packet switching network from the number of relay packet switching devices and the voice packet reception time, and further obtains the delay time from the delay time distribution. Let T _i be the time when the i-th voice packet was sent, and R _i be the time it was received.
R _i is expressed by the following equation (1).

R _i = T _i + D _c + Q _i (i = 1, 2, 3,...) ...(1) Here, D _c is the fixed amount of delay time in the packet switching network, and Q _i is the variable delay time that differs for each voice packet. Represents a certain delay time. The distribution of Q _i can be found from equation (1) by setting Q _i = R _i −T _i −D _c , but the receiving side cannot know the transmission time T _i of the voice packet and cannot directly determine Q _i . From the measurable reception time R _i |Q _i
−Q _i-1 |=|R _i −R _i-1 −T| (where i=2, 3,
..., and T represents the transmission time interval of voice packets), find the average value of the distribution of |Q _i -Q _i-1 , and then find the distribution of Q _i . The distribution function of Q _i is F(X)
Then, F(X) is the number of relay packet switching devices K
It is known that when the average transmission delay time per relay packet switching device is equal, it is generally expressed by the gamma distribution of equation (2).

F(X)=λ ^K・^X K- ¹・e ^- 〓 K/λ represents the mean value E _f of the distribution.

F(X) is uniquely determined by finding the average value E _f . On the other hand, if E _s is the average value of the distribution of |Q _i −Q _i-1 |, E _s is expressed by equation (3).

E _s = 1/λ _2k-2 〓 ^i=k-1 { _i C _K-1 (2K-1-i)/2 ⁱ } ...(3) Therefore, the relationship between E _f and E _s is (4) The formula becomes

E _f = K・E _s [ _2k-2 〓 ^i=k-1 { _i C _K-1 (2K-1-i)/2 ⁱ }] ^-1 ...(4) That is, on the receiving side |Q _i − The mean value of the distribution of Q _i-1 |
By measuring E _s , E _f is found by calculating equation (4), and furthermore, using equation (2), the distribution function F(X) of Q _i , that is, the distribution of transmission delay time is found.

Next, a method for determining the delay time from the distribution function F(X) of Q _i will be shown below. From equation (2), the probability P that the delay time is greater than or equal to D is expressed by equation (5).

P=e ^- 〓 ^D _k-1 〓 ^r=0 (λD) ^r / r〓 ...(5) Also, since P in equation (5) represents the packet loss rate, in order to keep the call quality constant, ( In formula 5), it is sufficient to find the delay time D corresponding to λ (=K/E _f ) so that P is constant. That is, the reception side sequentially measures the voice packet reception times, calculates the average value E _s of the difference between the reception times of consecutive voice packets, and uses equation (4) to calculate the average transmission delay time E _f in the packet switching network, that is, Find K/λ. Furthermore(5)
Using the formula, the delay time D is calculated from the above λ, the number K of relay packet switching devices, and the packet loss rate P.

The memory table of the delay time memory 18 shown in FIG. 5 stores the average value E _s (=E ₁ , E ₂ ,
..., E _M ) and the number of relay packet switching devices K (=1,
₂ , .

Therefore, as described above with reference to FIG. 6, the delay time is stored using the average value _Es of the difference in reception times of two consecutive voice packets in the n-th talk spurt and the number K of relay packet switching devices as parameters. The delay time D is read out from the unit 18, and furthermore, at the nth pause, it is confirmed that no audio packets are stored in the buffer memory 15 for absorbing delay time variation, and the delay time is updated to the delay time D. For the audio packet of the (n+1)th talk spurt, the newly updated delay time D absorbs the variation in transmission delay time. In the (n+1)th talk spurt, the first received audio packet is stored in the buffer memory 15 for absorbing delay time variation by the delay time D, and then the sequentially stored audio packets are stored in the first-in-first-out method. The audio packets are output to the audio packet output terminal 21 to absorb variations in transmission delay time of individual audio packets.

As explained above, in the first embodiment, the delay time for correcting the variation in the transmission delay time of voice packets is controlled in accordance with the transmission delay distribution in the packet switching network, thereby reducing the variation in the transmission delay time. It is possible to maintain the quality of the call regardless of the size of the transmission delay, and when the variation in transmission delay time is small, the amount of delay can be reduced, eliminating the drawback that the other party's voice is delayed and it becomes difficult to have a conversation. There are advantages that can be achieved.

In the first embodiment, the delay time storage unit 18 stores the average value E _s of the difference in reception time of each of two consecutive voice packets of a plurality of voice packets in one talk spurt and the delay corresponding to the number K of relay packet switching devices. By storing the time D _NM and controlling the delay time, fluctuations in the transmission delay time of individual voice packets were corrected. A similar effect can be obtained by storing the number of voice packets received from the received voice packets.

That is, the number of received voice packets X _NM (≒delay time D _NM / voice packet transmission time interval) corresponding to the delay time D _NM in the first embodiment is stored in the storage table shown in FIG. 5 as the average value _Es and is stored in correspondence with the number K of relay packet switching devices,
Similarly to the first embodiment, in the n-th talk spurt in FIG. 6, the delay time storage unit 1 uses the average value _Es and the number K of relay packet switching devices as parameters.
The received number X is read from 8. Next, at the n-th pause, the buffer memory 15 for absorbing delay time fluctuations
After confirming that no voice packets have been accumulated in , the number of received packets is updated to the number of received packets
For the +1st talk spurt, fluctuations in transmission delay time are absorbed by the newly updated reception number X. In the (n+1)th talk spurt, the first received voice packet is stored in the buffer memory 15 for absorbing delay time fluctuations until the voice packets following the first voice packet by the number of received voice packets are received. The sequentially accumulated voice packets are outputted to the voice packet output terminal 21 in a first-in-first-out manner, thereby absorbing variations in the transmission delay time of individual voice packets.

Since the present invention corrects variations in transmission delay time according to the distribution of transmission delays in the packet switching network, it is possible to maintain call quality regardless of the magnitude of variation in transmission delay time, and When the fluctuations in transmission delay time are small, the amount of delay can be reduced, which eliminates the disadvantage that the other party's voice is delayed and makes it difficult to have a conversation, providing a very effective transmission delay control method in voice packet switching systems. It is something to do.

[Brief explanation of drawings]

Figure 1 is a schematic diagram of a voice packet switching system.
Figure 2 is a distribution diagram of transmission delay time in a voice packet switching system, Figure 3 is an approximate distribution diagram of transmission delay time used in the conventional system, and Figure 4 is an embodiment of the present invention. 5 is an explanatory diagram showing the storage contents of the delay time storage section 18 of FIG. 4, and FIG. 6 is a time chart of the audio signal. 10...Delay time fluctuation absorbing device, 11...Relay packet switching device number input terminal, 12...Relay packet switching device number receiving circuit, 13...Processing device, 14...Audio packet input terminal, 15...Buffer memory for absorbing delay time variation, 16 ...Voice packet reception time measuring circuit, 17, 20...Timer circuit, 18...Delay time storage section, 19...Control circuit, 21...Voice packet output terminal.

Claims (1)

[Scope of Claims] 1. In a voice packet switching system that relays voice packets through one or more packet switching devices and transmits voice packets through the same route for each call, there is provided means for receiving the number of relay packet switching devices on a receiving side. , means for measuring the reception time of a voice packet, and means for determining the average value of the difference in reception time of two consecutive voice packets of the plurality of voice packets for each talk spurt (sound part) of the voice signal; means for storing a delay time determined according to a distribution of variations in transmission delay time of voice packets corresponding to the number of relay packet switching devices and the average value; a buffer memory for sequentially accumulating received voice packets; control means for extracting voice packets input to the buffer memory at regular intervals after accumulation for a specified time, reading out the delay time corresponding to the number of received relay packet switching devices and the average value from the storage means; After storing the first audio packet in each talk spurt (sound part) of the audio signal in the buffer memory for the read delay time, the fluctuations in the transmission delay time are absorbed by sequentially extracting the audio packets stored in the buffer memory. A voice packet transmission delay control method characterized by: 2. In a voice packet switching system in which voice packets are relayed through one or more packet switching devices and voice packets are transmitted through the same route for each call, there is a means for receiving the number of relay packet switching devices on the receiving side and a method for indicating the reception time of voice packets. means for determining the average value of the difference in reception time of two consecutive audio packets of a plurality of audio packets for each talk spurt (sound part) of the audio signal; and means for determining the number of relay packet switching devices. means for storing the number of received audio packets corresponding to a delay time determined according to a distribution of variations in transmission delay time of audio packets corresponding to the average value; and a buffer memory for sequentially storing received audio packets. , a control means for extracting voice packets inputted to the buffer memory at regular intervals after accumulation for a designated time, and a control means for storing the number of received voice packets corresponding to the number of received relay packet switching devices and the average value. , and store the first audio packet in each talk spurt (sound part) of the audio signal in the buffer memory until the number of audio packets read out is received continuously from the reception of the first audio packet, and then store the first audio packet in the buffer memory. A voice packet transmission delay control method characterized by absorbing variations in transmission delay time by sequentially extracting accumulated voice packets.