JP4269933B2 - IP telephone dynamic buffer control method and apparatus - Google Patents

Description

  The present invention relates to voice communication performed over an IP (Internet Protocol) network. More specifically, the present invention intends to provide a method and apparatus for dynamically controlling a buffer of an IP phone to enable high-quality calls.

    In voice packet transmission (hereinafter referred to as IP telephone) performed over an IP (Internet Protocol) network, fluctuations in the propagation time of voice packets (hereinafter referred to as packets) in the network have a significant effect on voice call quality. Therefore, a certain voice call quality is ensured by absorbing fluctuations in the buffer of the receiving terminal.

  There are a fixed buffer control system that keeps the buffer size of the receiving terminal constant, and a dynamic buffer control system that dynamically changes the buffer size using a RTP (Real Time Protocol) time stamp. In these receiving terminals, when a packet arrives from the IP network, it is temporarily stored in a buffer and arranged in the correct order to reproduce the voice. However, when the IP network is congested and packet loss occurs or the packet delay fluctuates, it becomes difficult to properly receive the packet in the buffer.

  In the dynamic buffer control method using a RTP (Real Time Protocol) time stamp, an abrupt change in jitter is instantaneously tracked based on information on fluctuations in packet reception intervals (hereinafter referred to as jitter). Such buffer increase / decrease processing is performed. However, since the buffer increase / decrease process itself causes a decrease in voice call quality, such buffer control in which the number of buffer increase / decrease processes frequently occurs is inappropriate for maintaining good call quality.

  FIG. 23 shows a buffer operation during normal reception of the conventional buffer unit. In (a), a Seq No. indicating the order of received packets and the reception timing thereof are indicated by bold lines. In (b), for example, a buffer having a buffer size of 5 (FIFO: First In First Out) of A to E included in the buffer unit and the stored Seq No. are shown. (C) shows the timing and Seq No. at which audio data is output. In (d), a reception clock having a period T (for example, 10 ms) in the voice receiving terminal is shown. The buffer unit stores the packets in the correct order according to the Seq No. of the received packet.

  The timing of the received packet in (a) and the timing of the received clock in (d) are the same, and the reception order of the received packets is not disturbed (replaced before and after) (the fluctuation of the propagation time in the IP network is within the allowable range). ). When a received packet of Seq No. 1 is received at t = 0, it is immediately stored in the buffer A. Similarly, when a received packet of Seq No. 2 is received at t = T, it is immediately stored in buffer A, and the audio data of Seq No. 1 is shifted to buffer B. This storing and shifting operation is executed within a sufficiently short time (for example, 10 ns) compared to the period T. When this operation is repeated until t = 4T, the five buffers A to E become full. Immediately after that, Seq No. 1 audio data is output. Thereafter, when a received packet of Seq No. 6 is received at t = 5T, immediately after that, the audio data of Seq No. 2 is output.

  FIG. 24 shows a buffer operation when an overrun of the conventional buffer unit occurs. The symbols shown in (a), (b), (c), and (d) are the same as those shown in FIG. However, (xx) written in the buffer (b) means an unused buffer for storing xx. The occurrence of the overrun (or underrun) tendency occurs when the influence of the fluctuation of the IP network is sufficiently small and can be ignored, when the clock period of the transmitting terminal is slightly shorter (or longer) than the clock period T of the receiving terminal. appear. In practice, it is impossible for both clock periods to coincide completely over a long period of time. In FIG. 24, for easy understanding, the period of the received packet (a) is exaggerated and shorter than the reception clock period T of (d).

  Until just before receiving the received packet Seq No. 9 in FIG. 24A, Seq No. 8 (7 to 4) is stored in the buffer A (B to E) in (b), but t = Shifted by 8T, Seq No. 9 can be written to buffer A. Therefore, when the received packet Seq No. 9 is received at the timing of t = 8T, it is written into the buffer A in (b) and the audio data output of Seq No. 5 is obtained in (c). The received packet in (a) has an overrun tendency (the time interval of the received packet is smaller than T), and when the Seq No. 10 arrives, the received packet Seq No. 10 can still be written to the buffer A. Not. Therefore, the received packet Seq No. 10 is discarded.

  At time t = 9T, the buffer A in (b) can write the received packet Seq No. 10, but remains unused, the data in Seq No. 9 to 6 is shifted, and (c) contains Seq. No.6 audio data output is obtained. When the received packet Seq No. 11 arrives (9T <t <10T), the buffer A can write the received packet Seq No. 10, and the Seq No. 11 is written and discarded without being left unused. Become. Similarly, when t = 13T, all the buffers A to E remain unused, and the audio data of Seq No. 10 is not obtained in (c), and the sound becomes silent. Accordingly, the silent state continues after t = 13T. Thus, once an overrun occurs, it cannot be received for all subsequent packets.

  In order to avoid this unreceivable situation, a buffer may be added to the future (above buffer A). However, if the overrun trend continues, more buffers must be added to the future side. That is, in order to cope with the semi-permanent continuation of the overrun tendency, it is necessary to provide a huge number of buffers on the future side from the beginning or to continue adding buffers to the future side. However, when a large delay occurs in the packet in the middle of overrun, the reception function may be temporarily recovered, but this is only temporary recovery and cannot be a countermeasure against overrun.

  FIG. 25 shows a buffer operation when an underrun occurs in the conventional buffer unit. The symbols shown in (a), (b), (c), and (d) are the same as those shown in FIGS. When the received packet Seq No. 9 in FIG. 25A is received at the timing of t = 8T, it is written in the buffer A in (b), and the audio data output of Seq No. 5 is obtained in (c). Shifted at the timing of t = 9T, Seq Nos. 9 to 6 are stored in buffers B to E, buffer A becomes an unused buffer waiting for the arrival of Seq No. 10, and (c) shows Seq No. 6 Audio data output is obtained. The received packet in (a) tends to underrun, and when Seq No. 10 arrives, the buffer A is already ready to write the received packet Seq No. 10, and Seq No. 10 is immediately written ( (The written state is not shown).

  Similarly, it is shifted at the timing of t = 10T, Seq Nos. 10 to 7 are stored in the buffers B to E, the buffer A becomes an unused buffer waiting for the arrival of Seq No. 11, and (c) includes Seq No.7 audio data output is obtained. When the received packet Seq No. 11 in (a) arrives, the buffer A is already in a state where the received packet Seq No. 11 can be written, and Seq No. 11 is written immediately (the written state is not shown). ).

In this way, when shifted at the timing of t = 13T, Seq Nos. 12 to 10 are stored in the buffers C to E, and the buffers A and B become unused buffers waiting for the arrival of Seq Nos. 14 and 13. , (C), Seq No. 10 audio data output is obtained. At the timing of t = 13T, the number of unused buffers has increased to two, A and B. As time passes, the number of unused buffers increases, and all buffers A to E become unused. Thereafter, the audio data output of (c) is not obtained at all, and reception is impossible.
JP 2002-64545 A JP 2002-185498 A

  The occurrence of an overrun (or underrun) tendency occurs when the clock period of the transmitting terminal is slightly shorter (or longer) than the clock period of the receiving terminal, ignoring the influence of fluctuations in the propagation time of the IP network. Actually, the clock periods of both the transmitting terminal and the receiving terminal cannot coincide completely over a long period of time. Therefore, overrun or underrun is a problem that always occurs. The buffer is initialized as a countermeasure when an overrun or underrun occurs. In this case, the call quality is significantly deteriorated. Furthermore, there may occur fluctuations in which the packet reception interval suddenly changes depending on the situation of the IP network. Immediate increase / decrease of the buffer in response to such fluctuation increases the frequency of buffer increase / decrease. Thus, the buffer increase / decrease process itself causes a significant decrease in voice call quality.

  Therefore, as a countermeasure when an overrun or underrun occurs, the buffer is not initialized, and even when fluctuations occur in the packet reception interval, the buffer is not immediately increased or decreased, or Preventing a significant drop in voice call quality without performing initialization is a problem to be solved. In addition, even when the network environment changes significantly, such as when the terminal is moved or when there is a significant change in traffic, there is no need to manually change the buffer size and it is stable without any manual operation. It is desired to obtain good maintainability for performing the above operation.

  The main feature of the present invention is to solve the above problems. Set up a run detection buffer to detect the tendency of overrun and underrun of the received data of the received packet among the multiple buffers that store the received data of the received packet, and analyze the reception position there Thus, the run tendency is detected, and when the run tendency is detected, the reception position is corrected. For sudden jitter, the received data loss rate of the received data in a predetermined period is measured, and if it is large, the number of buffers composed of a plurality of buffers is increased, and the received data delay packet When the rate was measured and it was small, the number of buffers consisting of the plurality of buffers was decreased to shorten the time from reception to audio data output.

  According to the dynamic buffer control method and apparatus of the IP telephone of the present invention, the tendency of overrun and underrun is detected for a predetermined period, and when the run tendency continues, the reception position and the position of the detection buffer are adjusted accordingly. Since the number of buffers was increased or decreased when the received data loss rate and the delayed packet rate were outside the predetermined range, there was no immediate response to sudden fluctuations in the packet reception interval, The frequency of buffer size changes has also been greatly reduced, preventing sudden fluctuations and failures due to overruns and underruns, improving voice call quality and at the same time achieving high maintainability.

  A reception data loss / delay rate checker for confirming the loss rate and delay packet rate of received data for a predetermined period, a run tendency detector for detecting a run tendency of received packets, and the confirmation data and run This was realized by providing a buffer increase / decrease unit to correct the reception position and the position of the run detection buffer, and a reception / run detection position correction unit to increase / decrease the buffer according to the trend.

  FIG. 1 is a block diagram of an embodiment of the present invention. The packet receiving unit 5 is connected to the IP network via the transmission path I / F unit 2 and, when receiving the packet, sends the received packet 15 to the dynamic buffer unit 6. The dynamic buffer unit 6 includes a plurality of buffers including a detection buffer. A reception position signal 16 indicating to which buffer the received packet 15 is received and a detection buffer signal 17 detected in the detection buffer are received as reception positions. The data is sent to the analysis / control unit 10.

  The reception position analysis / control unit 10 receives the reception position signal 16 and the detection buffer signal 17, analyzes the reception position of the reception packet 15 and the position of the run detection buffer, and receives a buffer increase / decrease instruction signal 38 and a reception / run command. A detection position instruction signal 39 is sent to the dynamic buffer unit 6. Under the control of the reception position analysis / control unit 10, the dynamic buffer unit 6 obtains audio data 19 and sends it to the audio encoding unit 7. The output of the speech encoding unit 7 is applied to the speech terminal via the terminal I / F unit 3. A voice signal from the voice terminal is applied to the packet assembly / transmission unit 8 via the terminal I / F unit 3, and its output is sent to the IP network side via the transmission line I / F unit 2.

  FIG. 2 shows the internal configuration of the reception position analysis / control unit 10 in relation to the dynamic buffer unit 6. The reception position analysis / control unit 10 includes a reception processing unit 20 and a reproduction processing unit 30. The reception processing unit 20 includes a position calculation data receiver 21, a reception data loss / delay rate checker 22, and a run tendency detector 23. The reproduction processing unit 30 includes a buffer increase / decrease unit 31 and a reception / run detection position corrector 32. Upon receiving the reception position signal 16 and the detection buffer signal 17, the position calculation data receiver 21 calculates the temporal position of the reception position signal 16 and the position of the detection buffer, and receives the reception position / run tendency signal 27 as the reception data loss / The data is sent to the delay rate checker 22 and the run tendency detector 23.

The reception data loss / delay rate checker 22 receives the reception position / run tendency signal 27 and obtains a reception data loss rate L _p that is a rate of loss of received packets and a delay packet rate D _D that is a rate of delay packets. Ask. The reception data loss rate L _p is obtained by equation (1). Here, T _p is a predetermined period, T is the period of the reception clock, and N _R is the number of packets received during the predetermined period T _p .
L _p = [{T _p / (TN _R )} − 1] × 100 [%] (1)
In the formula (1), for example, when T _p = 100T and N _R = 99, L _p is 1%. If {T _p / (TN _R )} <1, packet loss occurs (for example, T _p = 100T, N _R = 101). The packet loss occurs when the received packet 15 is received earlier due to fluctuations (jitter) in the IP network than when expected at the normal time. The delayed packet is generated when the received packet 15 is received at a point delayed by fluctuation (jitter) in the IP network from a point expected at normal time.

    FIG. 3 shows an operation of buffer reduction processing to cope with a case where a delay packet is suddenly generated and the delay packet rate, which is the generation rate, is reduced. In (a), the Seq No. indicating the order of the received packets 15 and the reception timing thereof are indicated by bold lines. In (b), buffers A to E included in the dynamic buffer unit 6, for example, a buffer size 5 (FIFO: First In First Out), and the Seq No. of the stored packet are shown. Yes. “E” and “D” indicate a delay detection buffer. It can be detected by monitoring the Seq No. of the delay detection buffer that the received packet 15 in (a) is delayed from the received clock in (d). (C) shows the timing at which the audio data 19 is output and the Seq No. of the packet. (D) shows a reception clock having a period T in the voice receiving terminal.

The delay detection buffer is set at the buffer positions E and D at the position closest to the audio reproduction (most delayed), and is directly received without being received by the other buffers A to D or A to C. A packet received by the detection buffers E and D is defined as a delay detection packet. Therefore, a packet that is received in the buffers A to D or A to C and then stored in the delay detection buffer E or D after the shift down is not a delay detection packet. Delayed packets ratio D _D is the percentage of the delay detection packet number N _D of the received packet number N _R received in a predetermined period of time T _p is as equation (2).
D _D = (N _D / N _R ) × 100 [%] (2)

Referring to the FIG. 3 this, for example, when setting the predetermined time period T _p in 9T～11T (actually longer period), in this period 9T~11T, Seq No of the received packet 15 (a) .10-12 are received (Seq No. 12 is equivalent to that received at t = 11T). The number of received packets is N _R = 2, and delay detection at each time point of t = 9T, t = 10T, and t = 11T The buffers are Chi E, Chi D, Chi D, and are stored in the delay detection buffers Chi E, Chi D directly without being received by AD (t = 9T) or AC (t = 10T, 11T). Since there is no received packet, the number of delay detection packets N _D = 0. Delayed packets rate _{D D} get 0%.

This delay packet rate D _D is determined in the received data loss deferral rate validator 22 similarly to the received data loss rate L _P. This delay packet rate D _D, if the value of the delay packet rate D _D is less than the value of the allowable delay packet rate D _Dth is a predetermined value, the buffer size (number of buffers) is too long than the appropriate size Therefore, the delay E buffer C is deleted at t = 9T + ΔT, that is, the number of buffers is reduced, and the buffer closest to the audio reproduction is set as D to be a delay detection buffer. Here, T is, for example, 10 ms, and ΔT is, for example, about 10 ns.

  FIG. 4 is an enlarged view of a part of the time axis of FIG. 3 illustrating the buffer reduction process. At t = 10T, the buffer A of the dynamic buffer unit 6 in (b) is an unused buffer because Seq No. 11 of the received packet 15 in (a) has not been received. When Seq No. 11 of the received packet 15 in (a) is received, the buffer A stores Seq No. 11. When t = 11T, the data in the dynamic buffer unit 6 in (b) is shifted, and the buffer A becomes an unused buffer because Seq No. 12 of the received packet 15 in (a) has not been received. When Seq No. 12 of received packet 15 in (a) is received, buffer A stores Seq No. 12. Here, Seq No. 11 of the received packet 15 of (a) is received between t = 10T and t = 11T, but at the time of t = 11T, it is the same as when received at the time of t = 10T. It is the result.

  FIG. 5 illustrates how the delay packet rate is calculated in the buffer reduction process of the dynamic buffer unit 6 when a received packet is received at a timing different from that in FIG. The dynamic buffer unit 6 includes buffers A to E, and E is a detection buffer. Before t = T, the buffers A to E of the dynamic buffer unit 6 in FIG. 6B remain empty and are unused buffers.

  When Seq No. 1 of the received packet 15 in FIG. 5A is received at t = T, it is stored in the buffer A. Seq Nos. 2 to 5 of the received packet 15 in (a) are received later than the normal time due to sudden fluctuation (jitter). From Seq No. 6 onward of the received packet 15 in (a), it is received at a normal timing. For example, Seq No. 2 of the received packet 15 in (a) can be received at t = 2T if normal. It is received at 3T ≦ t <4T due to sudden fluctuation, and the result is the same as when it is received at t = 3T, so that it is shown as received at t = 3T. Similarly, Seq No. 3 of the received packet 15 in (a) represents a case where 5T ≦ t <6T, Seq No. 4 is received with 5T ≦ t <6T, and Seq No. 5 is received with 9T ≦ t <10T. ing.

In FIG. 5, when a predetermined period T _p is set to 1T to 10T (T _p = 10T), Seq Nos. 1 to 10 of the received packet 15 of (a) are received (received packet) during this period 1T to 10T. The number N _R = 10). Only the packet of Seq No. 5 at t = 9T is directly stored in the delay detection buffer E of the dynamic buffer unit 6 in (b) without passing through other A to D (delay detection). Number of packets N _D = 1). Delay packet rate D _D = (N _D / N _R ) × 100 [%] = 10%. If the value of the small delay packet rate D _D than the predetermined delay packet rate D _Dth lasted a predetermined period of time, the buffer E as required, to the buffer decreasing process described in FIG. 3 and 4. By reducing the number of buffers in the dynamic buffer unit 6, the time from reception to reproduction of the audio data 19 is shortened, and the sound quality is improved.

Receive data loss rate L _p and the delay packet rate D _D obtained by the formula (1) and (2), the Los delay rate signal 28 the reproduction processing unit 30 is output from the received data loss deferral rate validator 22 as It is applied to the included buffer increase / decrease unit 31. The run tendency detector 23 receives the reception position / run tendency signal 27 and detects whether the received packet 15 is received by a buffer designated as a run detection buffer among a plurality of buffers included in the dynamic buffer unit 6. Thus, the run tendency is confirmed, and the reception position / run tendency confirmation signal 29 is output.

  The buffer increase / decrease unit 31 receives the loss / delay rate signal 28, determines whether the number of the plurality of buffers included in the dynamic buffer unit 6 should be increased or decreased, and sets and changes the detection buffer as necessary. Therefore, a buffer increase / decrease instruction signal 38 is sent to the dynamic buffer unit 6. The reception / run detection position corrector 32 receives the reception position / run tendency confirmation signal 29 and instructs the buffer / buffer in the plurality of buffers included in the dynamic buffer unit 6 to store the next reception data 15. A detection position instruction signal 39 is sent to the dynamic buffer unit 6.

FIG. 6 shows an operation of buffer increase processing for coping with a case where the reception data loss rate L _p is large. In (a), the Seq No. indicating the order of the received packets 15 and the reception timing thereof are indicated by bold lines. (B) includes, for example, buffers (FIFO: First In First Out) of A to E (F) included in the dynamic buffer unit 6, and Seq of the stored packet. No. is shown. (C) shows the timing at which the audio data 19 is output and the Seq No. of the packet. (D) shows a reception clock having a period T at the receiving terminal.

If the value of the reception data loss (packet loss) rate L _p is larger than the predetermined value L _{pth in} the predetermined period T _p before the time t = 9T of the reception clock in FIG. Immediately after time t = 9T (t = 9T + ΔT, where 0 <ΔT << T: for example, ΔT = 10 ns, T = 10 ms), the silent data is written to the past side of the buffer E. Increase processing for adding the buffer F is performed. At time t = 10T, the received packet 15 of Seq No. 11 of (a) received during time 9T <t <10T is written to the buffer A, shifted in the following order, 10 to B, 9 to C , 8 in D, 7 in E, silence is stored in F, and silence is output as audio data 19 in (c). In this way, the generation of unused buffers is prevented.

  FIG. 7 is an enlarged view of a part of the time axis of FIG. 6 for explaining the buffer increase processing. At t = 9T, the buffer A of the dynamic buffer unit 6 in (b) receives and stores Seq No. 10 of the received packet 15 in (a). At a time point 91 immediately after that, the buffer A becomes an unused buffer to which Seq No. 11 can be written. Further, at time 92, an increase process for adding a buffer F in which silence data is written is performed. In this increase processing, the buffer A is not used and the application of Seq No. 11 is awaited. When Seq No. 11 of the received packet 15 in (a) is received, the buffer A stores Seq No. 11. The buffer F stores silence. When t = 10T, the silent data in the buffer F is output as the audio data 19 in (c). At time 101 immediately after t = 10T, the stored data is shifted down, the buffer A is not used, and the application of Seq No. 12 is awaited. When Seq No. 12 of received packet 15 in (a) is received, buffer A stores Seq No. 12.

  FIG. 8 shows the buffer operation of the overrun correction process of the dynamic buffer unit 6. In (a), the Seq No. indicating the order of the received packets 15 and the reception timing thereof are indicated by bold lines. (B) includes, for example, a buffer of 5 + 1 buffer size (FIFO: First In First Out) included in the dynamic buffer unit 6 and a packet (voice data) stored therein. Seq No. is shown. “O” indicates that the buffer is an overrun detection buffer. It can be detected by monitoring the Seq No. of the overrun detection buffer that the received packet 15 has a cycle shorter than the reception clock (which tends to be overrun). (C) shows the timing at which the audio data 19 is output and the Seq No. of the packet. (D) shows a reception clock having a period T at the receiving terminal.

  A buffer A is provided on the future side (above A) of the buffer A as a buffer for overrun countermeasures. Even if Seq No. 10 of the received packet 15 of (a) which tends to be overrun is received, it can be stored in the buffer a for countermeasure against overrun. Since the data stored in the buffer a means that there is an overrun tendency, the data stored at the time of t = 9T + ΔT (0 <ΔT << T) is shifted down by one buffer, and A Stores Seq No. 10 data, B stores Seq No. 9 data,... E stores Seq No. 6 data. The buffer a is in a state where the data of Seq No. 11 can be stored. Upon receipt of Seq No. 11 of the received packet 15 of (a), the buffer a stores the data of Seq No. 11.

  When t = 10T, the data is shifted by one buffer, buffer A stores the data of Seq No. 11, outputs (c) audio data 19 of Seq No. 7, and the data of Seq No. 12 to buffer a Can be stored (unused buffer). By detecting that there is a write in the overrun detection buffer a which was an unused buffer (t = 9T), overrun correction processing becomes possible.

  FIG. 9 is an enlarged view of a part of the time axis of FIG. 8 illustrating the overrun correction process. At t = 9T, Seq No. 10 of the received packet 15 of (a) that tends to overrun is received and stored in the buffer a for overrun countermeasures. Since the fact that the data is stored in the buffer a means that there is an overrun tendency, the data stored at the time point 91 of t = 9T + ΔT (0 <ΔT << T) is shifted down by one buffer, A stores data of Seq No. 10, B stores data of Seq No. 9,... E stores Seq No. 6 data.

  The buffer a is in a state where the data of Seq No. 11 can be stored. Upon receipt of Seq No. 11 of the received packet 15 of (a), the buffer a stores the data of Seq No. 11. When t = 10T, the data is shifted by one buffer, the buffer A stores the data of Seq No. 11, and outputs (c) audio data 19 of Seq No. 7. The buffer a is in a state (unused buffer) in which data of Seq No. 12 can be stored. When Seq No. 12 of the received packet 15 in (a) is received, the buffer a stores the data of Seq No. 12. Thus, the overrun correction process can be performed by detecting that the overrun detection buffer a which was an unused buffer has been written (t = 9T, etc.).

  FIG. 10 shows an enlarged view of the time axis at the time of overrun correction processing when the received packet 15 is received at a timing considerably elapsed from the timing shown in FIGS. Here, correction processing is shown when two packets of Seq No. 51 and 52 of the received packet 15 of (a) are received during 49T ≦ t <50T. When Seq No. 50 of the received packet 15 of (a) is received, it is stored in the buffer a. When t = 49T, the buffer is shifted down and the buffer a becomes an unused buffer capable of receiving Seq No. 51, and the data of Seq No. 46 is output as the audio data 19 of (c). When Seq No. 51 of the received packet 15 of (a) is received, it is stored in the unused buffer a. At time 107 immediately after that, the buffer a is set as an unused buffer that can receive Seq No. 52 of the received packet 15 in FIG.

  Furthermore, since a predetermined time has elapsed since the overrun tendency has already been detected, the buffer A is not used so that it can receive Seq No. 53 by shifting down at time 108 following the operation at time 107. Set as a buffer. The buffer A is an unused buffer that can receive Seq No. 52. The data of Seq No. 51 to 48 is stored in the buffers B to E, and the data of Seq No. 47 is discarded. When Seq No. 52 of the received packet 15 of (a) is received, it is stored in the unused buffer A. Immediately after that, at time t = 50T, the data of Seq No. 48 is output as the audio data 19 of (c).

  FIG. 11 shows the buffer operation of the underrun correction process of the dynamic buffer unit 6. In (a), the Seq No. indicating the order of the received packets 15 and the reception timing thereof are indicated by bold lines. In (b), for example, a buffer size 5 (FIFO: First In First Out) of A to E included in the dynamic buffer unit 6 and the Seq No. of the stored packet (voice data) are shown. It is shown. “A” indicates that the buffer B is an underrun detection buffer. It can be detected by monitoring the Seq No. of the underrun detection buffer that the received packet 15 comes in a cycle longer than the reception clock (which tends to underrun). (C) shows the output timing of the audio data 19 and the Seq No. of the packet (audio data). (D) shows a reception clock having a period T at the receiving terminal.

  Buffer B is provided on the past side of buffer A (below A) as a buffer for countermeasures against underrun. When receiving the received packet 15 of (a) that tends to underrun, it may not be possible to receive any of the received packets 15 of (a) during one period T of the received clock of (d). As a result, unused buffers (xx) may occur continuously. At the time point t = 10T when this has continued for a predetermined number of times, the buffers A and B are unused. After that, Seq No. 10 is directly received by Buffer B without going through Buffer A, Buffer A becomes an unused buffer that can store Seq No. 11 data, and Buffer A B stores Seq No. 10 data. , ... E stores silent data. When t = 11T, silence is output as the audio data 19 of (c). The buffer A remains an unused buffer that can store the data of Seq No. 11. Thereafter, upon receipt of the received packet 15 of Seq No. 11 (a), the buffer A can store the data of Seq No. 11. This underrun correction process can prevent an increase in unused buffers.

  FIG. 12 is an enlarged view of a part of the time axis of FIG. 11 for explaining the underrun correction processing. Before t = 10T, it keeps monitoring that the buffer A is always unused, and detects an underrun tendency. At t = 10T, buffers A and B are unused and the underrun trend continues. In response to Seq No. 10 of the received packet 15 of (a), the buffer of the underrun detection buffer B or the past side (buffers C, D, E) without passing through the buffer A and the underrun detection buffer B Store directly in C. The fact that the direct reception in the buffer C means that the underrun tendency has continued for a long period of time, so that at the time 105 immediately after receiving the Seq No. 10 of the received packet 15, the silent data is stored in the buffer E near the output side Store. When t = 11T, (c) silence is output as the audio data 19, and the buffer A becomes an unused buffer capable of storing Seq No. 11 data. Thereafter, upon receipt of the received packet 15 of (a) of Seq No. 11, the buffer A stores the data of Seq No. 11. This underrun correction process prevents an increase in unused buffers.

  13 and 14 show the flow of operations of the reception processing unit 20 included in the reception position analysis / control unit 10. When the operation starts, all the buffers of the dynamic buffer unit 6 are emptied so that the buffer A can receive the received packet 15 of Seq No. 1. That is, the buffer A is set to an unused buffer (1) and temporarily initialized (S11, FIG. 13). Waiting to receive the received packet 15 in the buffer (S12), receiving the received packet 15 and detecting a reception interrupt (S13Y). The reception interruption is the first reception after the start of the call or the first reception after the start of the subsequent call because a part of the received packet 15 is lost due to the fluctuation of the propagation time in the IP network or the run tendency. (S14Y), the buffer is initialized and an unused buffer (xx) is prepared for the Seq No. of the received packet 15 to be received next (S15).

If Seq No. N of received packet 15 is not the first received after the start of the call (S14N) or if it is initialized (S15), is Seq No. It is determined whether it is a storage object or a future object that has not yet been stored (S16, FIG. 14). For the past, because not in the already stored object, the discarding of the received data (received packet 15) (S17), counting the data discard number N _A (S18). By counting the data discard number N _A, it is possible to know the number of processing unnecessary data.

  If it is to be stored, the data is written in the dynamic buffer unit 6 and the Seq No. is updated (S19). If the Seq No. of the received packet 15 is in the future than the storage target (a part of the received packet is missing), the dynamic buffer 6 is reconstructed and the received packet 15 of the subsequent Seq No. An unused buffer (xx) is set so that it can be stored (S20).

When the position calculation data receiver 21 confirms that the data received by the reception position signal 16 is position calculation data (S21Y), a process of checking the reception data loss rate L _p of the reception packet 15 is performed (S22). ) The packet reception position is confirmed, and the operation returns to the operation of step S12 (FIG. 13) (S23).

FIG. 15 shows a subroutine of the data loss rate confirmation processing in step S22 of FIG. When the data loss rate confirmation processing is started, the received data loss rate L _p is the timing for analyzing the received data loss rate L _p (S31Y), and the received data discard number N _A and the received packet number shown at time 108 in FIG. From N _R , the received data loss rate L _p and the delayed packet rate D _D are calculated by Equations (1) and (2).

When the reception data loss rate L _p is compared with a predetermined reception data loss rate L _pth and L _p ≧ L _pth is satisfied (S33Y), the buffer increase bit (Incbit) is used to perform the buffer increase processing shown in FIG. ) was turned on (oN) (S34), the data discard number _{N a} is reset to 0 (S35), the received data loss rate confirmation process (S22, FIG. 14) ends. The reception data loss rate confirmation process (S22, FIG. 14) ends even when it is not time to analyze (S31N) and L _p ≧ L _pth is not satisfied (S33N).

FIGS. 16 and 17 show a subroutine of the packet reception position confirmation process in step S23 of FIG. When the packet reception position confirmation process is started, for example, at which position X _i of the buffer in FIG. 11 the received packet 15 is received (X _i = A B in FIG. 11) is calculated based on the reception position. (S51, FIG. 16).

By receiving the position X _i, when confirming that it is time to analyze (S52Y), the buffer reduces determination processing coefficients β calculating a [%]. For example, at t = 9T of the dynamic buffer unit 6 in FIG. 3B, some of the 100 received packets 15 are directly received by the delay detection buffer E without passing through the other buffers A to D. For example, when two are detected, β = 2% is calculated. The buffer decrease determination processing coefficient β is compared with a predetermined buffer decrease determination processing coefficient β _th (for example, 3%). If β <β _th (S54Y), the buffer decrease processing is performed to perform the buffer decrease processing shown in FIG. The decrease bit (Decbit) is turned on (S55), and the process proceeds to the overrun detection step S56 (FIG. 17). Step S54 may not be beta <beta _th (FIG. 14) proceeds to (S54N), the overrun detection step S56 (FIG. 17).

  When an overrun is detected by the overrun detection buffer (see FIG. 8) (S56Y, FIG. 17), the overbit (Overbit) is turned on (ON in FIG. 8 is added on A). (S57). When the underrun is detected (S56N, S58Y), the underbit (Underbit) is turned on (ON) (silence data is inserted on the most past side at t = 10T + ΔT in FIG. 11) (S59). Therefore, in preparation for the next time, the period m for analyzing the reception position is incremented to m + 1 (S60), the index i of the reception position is incremented to i + 1, and the packet reception position confirmation processing subroutine ends (S61). ). If no underrun is detected (S58N), the process proceeds to step S60. Even if it is not time to analyze (S52N, FIG. 16), the process proceeds to step S61.

  FIG. 18 shows a flow of operations of the reproduction processing unit 30 included in the reception position analysis / control unit 10. When the reproduction interruption is detected (S81Y), the oldest stored data in the buffer is reproduced as the audio data 19 of (c) of FIG. 3 (S82). Therefore, BFI (provisional number) data having a Seq No. that is equal to the Seq No. of the latest stored data is set as an unused buffer (xx) in the buffer (S83).

  Based on the results of overrun detection (S57, FIG. 17) and underrun detection (S59, FIG. 17), the necessity of position correction of the received packet is examined (S84). When overrun is detected and Overbit is ON, the overrun correction process is started (S85). If underrun is detected and Underrbit ON, underrun correction processing is entered (S86). The necessity of buffer increase / decrease processing is examined when each correction processing is required and when it is not required (S87). If buffer increase is required (FIG. 6), buffer increase processing is performed (S88). If buffer reduction is required (FIG. 4), buffer reduction processing is performed (S89). After performing the buffer increase / decrease process and when it is not necessary to return to the operation in step S83.

  FIG. 19 shows a subroutine of overrun correction processing (S85, FIG. 18). When the overrun correction process is started, the silence detection bit is ON (silence data is written in the buffer E of t = 10T + ΔT in FIG. 11), or the tendency of overrun or underrun is a predetermined time. It is checked whether or not the timer bit 1 indicating the continuation is ON (S101). If neither is ON, overrun correction processing is not required (S101N).

  If the silence detection bit or timer bit 1 is ON (S101Y), the oldest packet in the buffer (FIG. 8, t = 9T buffer E) is discarded and converted to BFI data (S102). . That is, the data contents of the buffers a and A to E at t = 9T + ΔT are changed from the data contents of the buffers a and A to E at t = 9T in FIG. Therefore, the first packet after the BFI data conversion (the received packet 15 of Seq No. 11 in FIG. 8) is rearranged at the optimal reception position (buffer A at t = 10T in FIG. 8), and other voice packets are also appropriate. Relocation to the position (as in buffers B to E at t = 10T in FIG. 8) (S103). When the rearrangement is completed, the silence detection bit is set to 0 (the state of t = 11T in FIG. 11), the timer bit 1 is reset to 0 (the start of the run tendency monitoring period), and the over bit is set to 0 (overrun The state is reset to t = 9T in FIG. 8 before the correction processing, and the overrun correction processing subroutine is terminated (S104).

  FIG. 20 shows a subroutine of the underrun correction process (S86, FIG. 18). When the underrun correction process is started, the silence detection bit is ON (silence data is written in the buffer E of t = 10T + ΔT in FIG. 11), or the tendency of overrun or underrun is a predetermined time. It is checked whether or not the timer bit 1 indicating the continuation is ON (S111). If neither is ON, underrun correction processing is not required (S111N).

  When the silence detection bit or timer bit 1 is ON (S111Y), the first packet (the received packet 15 of Seq No. 10 in FIG. 11) is the buffer at the optimum reception position (t = 10T in FIG. 11). B from buffer B to buffer B at t = 10T + ΔT) (position correction), and other voiced and silent packets are also rearranged to appropriate positions (as in buffers C to E at t = 10T + ΔT in FIG. 11). (S112). When the rearrangement (position correction) is completed, the silence detection bit is reset to 0 (the state of t = 11T in FIG. 11), timer bit 1 is set to 0 (starting the run trend monitoring period), and the over bit is set Reset to 0 (the state of t = 9T in FIG. 11 before the overrun correction process), and the underrun correction process subroutine ends (S113).

  FIG. 21 shows a subroutine of buffer increase processing (S88, FIG. 18). When the buffer increase process is started, when the buffer increase process in which the silence detection bit or the timer bit 2 is ON is required (S121Y), t = 9T + ΔT (0 <ΔT <in FIG. 6). As illustrated in the buffer F in <T), the buffer is increased by one packet (S122). Therefore, the silence detection bit is reset to 0 (the state of t = 10T in FIG. 6), the timer bit 2 is reset to 0 (to start the run tendency monitoring period), and the over bit is set to 0 (overrun correction processing) 6 (the state of t = 9T in FIG. 6) before, and the buffer increase processing subroutine ends (S123). If the buffer increase process is not required (S121N), the buffer increase process subroutine is immediately terminated.

  FIG. 22 shows a subroutine of buffer reduction processing (S89, FIG. 18). When the buffer reduction process is started, if the silence detection bit or timer bit 2 is ON and the buffer reduction process is required (S131Y), the buffer E at + ΔT at t = 9T in FIG. As illustrated in (1), the buffer is decreased by one packet (S132). Therefore, the silence detection bit is reset to 0, the timer bit 2 is reset to 0 (start of the run tendency monitoring period), the over bit is reset to 0, and the buffer reduction processing subroutine is terminated (S133). ). If the buffer reduction processing is not required (S131N), the buffer reduction processing subroutine is immediately terminated.

It is the block diagram which showed the Example of the dynamic buffer control apparatus of IP telephone. Example 1 FIG. 2 is a circuit configuration diagram of a reception position analysis / control unit that is a component of FIG. FIG. 2 is a buffer operation diagram at the time of buffer reduction processing of a dynamic buffer unit that is a component of FIG. 1; FIG. 4 is an enlarged view of the time axis of FIG. FIG. 4 is a buffer operation diagram for explaining delay packet rate calculation at the time of buffer reduction processing of the dynamic buffer unit when a received packet is received at a timing different from that in FIG. FIG. 2 is a buffer operation diagram at the time of buffer increase processing of a dynamic buffer unit that is a component of FIG. FIG. 7 is an enlarged view of the time axis of FIG. FIG. 2 is a buffer operation diagram when an overrun tendency occurs in a dynamic buffer unit that is a component of FIG. 1; FIG. 9 is an enlarged view of the time axis of FIG. FIG. 10 is an enlarged view of a time axis at the time of overrun correction processing when a received packet is received at a timing considerably after the timing shown in FIGS. 8 and 9. FIG. 2 is a buffer operation diagram when an underrun tendency occurs in the dynamic buffer unit that is a component of FIG. 1; FIG. 12 is an enlarged view of the time axis of FIG. 3 is a flowchart showing a flow of reception processing operation of a reception position analysis / control unit which is a component of FIG. 14 is a flowchart showing a flow of a reception processing operation of a reception position analysis / control unit which is a component of FIG. 1 together with FIG. FIG. 15 is a flowchart showing a flow of operation of a data loss rate confirmation process that is a subroutine of FIG. 14. FIG. 15 is a flowchart showing a flow of operation of packet reception position confirmation processing that is a subroutine of FIG. FIG. 15 is a flowchart showing a flow of operation of received data / loss rate confirmation processing which is a subroutine of FIG. 14 together with FIG. 16; 2 is a flowchart showing a flow of a reproduction processing operation of a reception position analysis / control unit which is a component of FIG. FIG. 19 is a flowchart showing a flow of an overrun correction processing operation that is a subroutine of FIG. 18. FIG. FIG. 19 is a flowchart showing a flow of an underrun correction processing operation that is a subroutine of FIG. 18. FIG. FIG. 19 is a flowchart showing a buffer increase processing operation that is a subroutine of FIG. 18; FIG. FIG. 19 is a flowchart showing a flow of buffer reduction processing operation that is a subroutine of FIG. 18; FIG. It is a buffer operation | movement figure at the time of the normal reception of the conventional buffer part. It is a buffer operation | movement figure at the time of the overrun occurrence of the conventional buffer part. It is a buffer operation | movement figure at the time of the underrun occurrence of the conventional buffer part.

Explanation of symbols

2 Transmission path I / F unit 3 Terminal I / F unit 5 Packet receiving unit 6 Dynamic buffer unit 7 Voice decoding unit 8 Packet assembly and transmission unit
DESCRIPTION OF SYMBOLS 10 Reception position analysis control part 15 Reception packet 16 Reception position signal 17 Detection buffer signal 19 Voice data 20 Reception processing part 21 Data receiver for position calculation 22 Received data loss / delay rate checker 23 Run tendency detector 27 Reception position / run Trend signal 28 Loss / delay rate signal 29 Reception position / run tendency confirmation signal 30 Playback processing unit 31 Buffer increase / decrease unit 32 Reception / run detection position corrector 38 Buffer increase / decrease instruction signal 39 Reception / run detection position instruction signal
91,92,101,105,107,108 point β buffer decrease determination processing coefficients D _D Seq No. delay packet rate D _Dth allowable delay packet index L _P number N received data in the received data loss rate i buffer (sequence number)
N _A Number of discarded data N _D Number of delay detection packets N _R Number of received packets T Period of received clock _TP Predetermined period

Claims (5)

Reception processing (10, 20) for analyzing reception positions in a plurality of buffers in which received data of received packets are stored,
When the received data loss rate (L _P ) of the received data in a predetermined period (T _P ) is large, the number of the plurality of buffers is increased, and the delayed packet rate (D _D ) of the received data is small Includes a reproduction process (10, 30) that enables reproduction of audio data by reducing the number of the plurality of buffers. In the reception process (10, 20), a run tendency detection process (23) for detecting at least one of the run tendency of the received data of the received packet from the reception position. And
When the run tendency is detected, in the reproduction process (10, 30), a reception position correction process (32) for correcting the reception position is performed.
2. The dynamic buffer control method for an IP phone according to claim 1. 3. The run detection buffer position correction process (32) for correcting a position of a run detection buffer for detecting the run tendency among the plurality of buffers in the reproduction process (10, 30). IP phone dynamic buffer control method. Receiving means (10, 20) for analyzing reception positions in a plurality of buffers in which received data of received packets are stored;
When the received data loss rate (L _P ) of the received data in a predetermined period (T _P ) is high, the number of the plurality of buffers is increased, and the delayed packet rate (D _D ) of the received data is small Includes a reproducing means (10, 30) for reproducing audio data so as to reduce the number of the plurality of buffers. The receiving means (10, 20) includes a run tendency detecting means (23) for detecting at least one run tendency of the received data of the received packet from the reception position. Including
When the run tendency is detected, the reproduction means (10, 30) includes reception position correction means (32) for correcting the reception position.
5. The dynamic buffer control device for an IP telephone according to claim 4.

Images