JP4635880B2 - Packet communication apparatus and packet communication method - Google Patents

Description

  The present invention relates to a packet communication apparatus and a packet communication method, and more particularly to a packet communication apparatus and a packet communication method for transmitting and receiving packetized voice data such as a so-called IP phone.

  A packet communication apparatus that transmits and receives packetized audio data includes a reception buffer and a jitter buffer. The reception buffer is a timing buffer that temporarily stores a predetermined number of received packets. The amount of control in this reception buffer is the number of received packets, and how to reduce the number of received packets discarded is an important issue. The jitter buffer is a buffer for preventing fluctuations for buffering and leveling voice data (packets) arriving within a certain time to prevent deterioration in sound quality due to fluctuations in arrival time. The amount of control in this jitter buffer is the buffer size (jitter buffer amount) necessary to level out fluctuations in arrival time and maintain high quality, but there is a trade-off between fluctuation leveling and voice delay. Therefore, an important issue is how to determine a buffer size that can satisfy both at the same time in consideration of the balance between fluctuation leveling and audio delay.

  Hereinafter, a conventional method for controlling the control amounts of these buffers will be described. FIG. 5 is a block diagram illustrating a configuration example of a conventional packet communication apparatus. A conventional packet communication apparatus 500 shown in FIG. 5 includes a packet reception control unit 502 that controls reception of a packet transmitted from a packet transmission source based on a packet communication protocol via a network 501, and a packet reception control unit 502. A reception buffer 503 for buffering the reception packet sent from the reception buffer, a reception buffer amount measurement unit 504 for measuring the number of reception packets stored in the reception buffer 503 and detecting the presence or absence of an overflow, and a predetermined fixed time The above-described jitter buffer 505 for buffering received packets arriving from the reception buffer 503 and leveling fluctuations in arrival time to maintain high quality, and control for transmitting output packets from the jitter buffer 505 to an external packet receiving device 507 A packet transmission control unit 506 To have.

  The changing unit 508 included in the packet reception control unit 502 has a function of changing the operation of the packet reception control unit 502 in response to the output (packet reception stop request) of the reception buffer amount measurement unit 504. Further, the changing means 509 provided in the reception buffer 503 has a function that allows the threshold value for the number of received packets to be buffered to be changed from the outside. Further, the changing means 510 provided in the jitter buffer 505 has a function of making it possible to change the buffer size from the outside.

  In the above configuration, first, a method for controlling the number of received packets, which is the control amount of the reception buffer 503, will be described with reference to FIGS. 5, 6, and (Table 1). FIG. 6 is a diagram for explaining the configuration and control operation principle of the reception buffer of FIG. Table 1 shows the relationship between the threshold value and the number of packets that can be stored in the reception buffer 503.

  In FIG. 5, the packet reception control unit 502 outputs a packet sent from the network 501 to the reception buffer 503. The reception buffer 503 is a FIFO (first-in first-out) buffer as shown in FIG. That is, in the reception buffer 503, the reception packets from the packet reception control unit 502 are sequentially shifted from the reception port 601 toward the transmission port 602, and packets that have reached the transmission port 602 are transmitted via the jitter buffer 505. The packet is transmitted to the packet receiving device 507 under the control of the control unit 506.

  In this case, when the speed of the received packet input from the packet reception control unit 502 is faster than the capturing speed of the packet receiving device 507, the received buffer stays and accumulates in the reception buffer 503, and thus overflow is detected. Therefore, a threshold value 604 that gives a limit value of the number of packets 603 to be stored is determined near the reception port 601. This threshold value 604 can be changed from the outside by using the changing means 509. For example, as shown in (Table 1), the threshold value 604 for detecting an overflow is a limit value (reception value) of the number of packets 603 accumulated. The number of possible packets is determined to be 900 kbytes.

  The reception buffer amount measurement unit 504 measures the number of packets 603 to be accumulated, and detects an overflow when the threshold value 604 is exceeded. In the example shown in FIG. 6, when the packet 606 is captured from the reception port 601 before the packet 605 existing at the transmission port 602 is transmitted in a situation where the number of accumulated packets 603 matches the threshold value 604, Detect overflow. When the reception buffer amount measurement unit 504 detects an overflow, the reception buffer amount measurement unit 504 causes the packet reception control unit 502 to execute processing 607 for requesting the communication partner (transmission source) to stop transmission.

  That is, when the reception buffer amount measurement unit 504 detects an overflow, the reception buffer amount measurement unit 504 outputs a packet reception stop request to the changing unit 508 provided in the packet reception control unit 502. In the packet reception control unit 502, the changing unit 508 receives a packet reception stop request from the reception buffer amount measurement unit 504 and creates a packet transmission stop message, so that the packet transmission stop message is sent to the packet transmission source via the network 501. Send to. At the same time, packet transmission to the reception buffer 503 is stopped according to the stop instruction created by the changing means 508.

  As a result, in the packet communication apparatus 500, packets stored in the reception buffer 503 are transmitted via the jitter buffer 505 and the packet transmission control unit 506 within a period in which the transmission source of the packet that is the communication partner stops packet transmission. Thus, the packet can be transmitted to the external packet receiving device 507, and the amount of packet discard due to overflow of the reception buffer 503 can be reduced.

  Next, a method for controlling the buffer size, which is the control amount of the jitter buffer 505 shown in FIG. 5, will be described with reference to FIGS. 7A and 7B are diagrams for explaining the control operation principle of the jitter buffer in FIG. 5. FIG. 7A illustrates a case where the jitter buffer size is small, and FIG. 7B illustrates a case where the jitter buffer size is large. To do. Table 2 shows an example of setting the buffer size, and Table 3 shows the relationship between the buffer size and various qualities.

  The jitter buffer 505 shown in FIG. 5 waits for the arrival of a packet for the buffering time, buffers the received packet, and sends out the packet from the buffer at a fixed interval. Therefore, the buffer size (jitter buffer capacity) Can be adjusted by setting the buffering time. This can be changed from the outside using the changing means 510 shown in FIG.

  FIG. 7 shows (a) when the jitter buffer size is small and (b) when the jitter buffer size is large. Here, as shown in Table 2, the buffering time when the buffer size is small is 5 packet hours, and the buffering time when the buffer size is large is 10 packet times.

  In FIG. 7A, the time that the jitter buffer (small) 701 can be buffered waits for reception and input of a packet from the receiving port 703 for 5 packet times after the first packet 702 is received. After the set time of 5 packet times has elapsed, the reception packet is transmitted from the transmission port 704 to the packet reception device 507 via the packet transmission control unit 506 at regular intervals.

  In FIG. 7B, the time that the jitter buffer (large) 711 can be buffered waits for reception and input of a packet from the receiving port 713 for 10 packet times after the first packet 712 is received. After the set time of 10 packet time has elapsed, the reception packet is transmitted from the transmission port 714 to the packet reception device 507 via the packet transmission control unit 506 at a constant interval.

  Here, since the maximum number of packets that can be buffered by the jitter buffer (small) 701 is 5 packet times, which is smaller than that of the jitter buffer (large) 711, the buffering delay in the jitter buffer (small) 701 is relatively small. I'll do it. However, when the jitter buffer capacity is small, a packet with a large arrival delay is treated as a loss, and therefore, when the packet is decoded, the loss of data is caused and the communication quality is greatly affected.

  On the other hand, since the number of packets that can be buffered by the jitter buffer (large) 711 having a large jitter buffer capacity is as large as 10 packet hours, the buffering delay in the jitter buffer (large) 711 is the jitter buffer (small) 701. Doubled compared to. However, since the jitter buffer capacity is large, a time margin for waiting for an arrival delay packet can be obtained. Therefore, data loss due to delay is less likely to occur compared to the jitter buffer (small) 701, and stable communication quality can be maintained when decoding a packet.

  Thus, the capacity of the jitter buffer has a trade-off relationship between “communication quality” and “buffering delay”. In particular, when exchanging audio data in packets, the “buffering delay” when decoding as audio is an important factor in “communication quality” of audio communication. A difficult setting balance of “call quality” requiring both “quality” and “buffering delay” is required. This relationship is shown in (Table 3).

As can be understood from the above explanation, when voice is packetized and communication is performed, the overflow control of the reception buffer and the adjustment of the jitter buffer amount have a very close relationship in forming the voice call quality. ing.
JP 2005-39719 A

  However, the conventional reception buffer overflow control mechanism is not fuzzy logic with ambiguity, but crisp logic that determines whether or not to perform overflow control depending on whether a threshold value is exceeded. Therefore, once the overflow control threshold is exceeded, rapid overflow control is performed. In other words, when communication is performed by packetizing voice, smooth voice quality is maintained up to the vicinity of the threshold, but once overflow control is performed, packet transmission and reception stop due to overflow control occur, and higher communication There is a problem that a procedure for data transmission and retransmission according to the protocol occurs, and the delay of packet data arriving at the jitter buffer increases abruptly, resulting in a sudden deterioration in call quality.

  This will be specifically described. Crisp logic is a clear binary logic whether the value belongs to a certain set or not, and has no ambiguity in the boundary, so that it is difficult to make an intermediate determination. For this reason, even when the number of received packets is slightly different in the vicinity of the threshold value for executing the overflow control or when it is far from the threshold value, there is no change in the application of the logic for executing the overflow control. That is, the case shown in (Table 4) occurs.

  (Table 4) shows the relationship between the number of received packets per unit time and the presence / absence of overflow control when the threshold value for performing overflow control is set to the number of received packets per unit time = 50. Like pattern # 1 with 51 received packets per unit time and pattern # 3 with 49 received packets per unit time, the number of received packets per unit time is a little across the threshold (= 50). Even if only two packets are different, the pattern # 1 with 51 received packets per unit time is the same as the pattern # 2 with 100 received packets per unit time with a large difference from the threshold (= 50). Overflow control logic is applied. Therefore, conventionally, as described above, when voice is packetized and communication is performed, there is an effect such as a sudden change in call quality near the threshold value for overflow control.

  In addition, since the jitter buffer capacity is not dynamically associated with changes such as an increase in packet arrival delay, the jitter buffer capacity increases with the sudden increase in packet arrival delay as a result of sudden overflow control. There is also a problem that it is difficult to respond dynamically and it is difficult to obtain a favorable call quality. The conventional overflow control described above is hereinafter simply referred to as “flow control”.

  In addition, regarding a mechanism for selecting a communication channel in a communication base station, an example has been proposed in which fuzzy control is applied to enable selection of an optimal communication channel at all times (see, for example, Patent Document 1).

  The present invention has been made in view of the above example, and enables the control amounts of the reception buffer and the jitter buffer to be adaptively controlled using fuzzy logic when packetized voice data is transmitted and received. Accordingly, an object of the present invention is to obtain a packet communication apparatus and a packet communication method that can always provide optimum call quality.

  In order to achieve the above object, a packet communication apparatus according to the present invention includes: a received packet number measuring unit that measures the number of packets received by a packet reception control unit that controls transmission / reception of packets with a transmission source; and a packet reception control unit. Reception buffer measurement means for measuring the number of packets stored in the reception buffer for temporarily storing input packets, and arrival variations of packets arriving from the reception buffer in a jitter buffer for leveling delay fluctuations of packets input from the reception buffer A jitter buffer measuring means for measuring an average time; a transmission packet number measuring means for measuring the number of packets transmitted by a packet transmission control means for controlling transmission of a packet stored in the jitter buffer to a receiver; a received packet number measuring means; Reception buffer measurement means, jitter buffer measurement means, and transmission packet Measurement information collecting means for collecting each measurement result of the number measuring means, a fuzzy inference engine unit for executing a fuzzy inference by applying a predetermined fuzzy inference rule to each measurement result collected by the measurement information collecting means, and a fuzzy inference Based on the fuzzy inference result of the engine unit, the reception control operation of the packet reception control means is changed, the setting of each buffer amount of the reception buffer and the jitter buffer is changed, and the transmission control operation to the reception source of the packet transmission control means is changed. And a changing means.

  According to the present invention, fuzzy threshold judgment used when changing the control amount of the conventional flow control or changing the amount of jitter buffer is input, and a plurality of inputs of flow control and control of the amount of jitter buffer are input. It can be performed in an optimal state by comprehensively judging information with different scales, and correspondingly changes the control operation of packet transmission / reception with the transmission source and reception source, so a slight reception across the threshold Without frequently performing flow control due to changes in the packet amount, it is possible to dynamically set and control the appropriate reception buffer and jitter buffer control amount (buffer amount) at all times. Quality can be maintained optimally.

  According to the present invention, when packetized voice data is transmitted / received, there is an effect that it is possible to always provide optimum call quality.

  Exemplary embodiments of a packet communication device and a packet communication method according to the present invention are explained in detail below with reference to the drawings. FIG. 1 is a block diagram of a packet communication apparatus according to an embodiment of the present invention. In FIG. 1, components that are the same as or equivalent to the components shown in FIG. In the following description, the part related to the present embodiment will be mainly described. Hereinafter, the configuration and operation of the fuzzy inference engine unit 105 will be described with reference to the drawings.

  The packet communication apparatus 100 shown in FIG. 1 has the configuration shown in FIG. 5, the received packet number measurement unit 101, the reception buffer measurement unit 102 instead of the reception buffer amount measurement unit 504, the jitter buffer measurement unit 103, and the transmission A packet number measurement unit 104, a fuzzy inference engine unit 105, a fuzzy inference rule change unit 106, an information input unit 107, and a user setting interface (user setting I / F) 108 are added. A user setting terminal 109 is connected to the user setting I / F 108.

  The packet reception control unit 502 is provided with a changing unit 110 instead of the changing unit 508, the reception buffer 503 is provided with a changing unit 111 instead of the changing unit 509, and the jitter buffer 505 is replaced with the changing unit 510. The changing means 112 is provided.

  The received packet number measuring unit 101 measures the number of packets received from the network 501 by the packet reception control unit 502. The reception buffer measurement unit 102 measures the number of packets stored in the reception buffer 503. The jitter buffer measuring unit 103 measures the arrival variation average time of packets arriving at the jitter buffer 505 from the reception buffer 503. Specifically, the jitter buffer measurement unit 103 measures a reception interval of packets received within a predetermined time and performs a process of obtaining an average value thereof. The transmission packet number measuring unit 104 measures the number of packets transmitted from the jitter buffer 505 toward the packet receiving device 507.

  The information input unit 107 includes the number of transmitted / received packets measured by the received packet number measuring unit 101 and the transmitted packet number measuring unit 104, the increase / decrease amount of packets derived from the number of packets measured by the reception buffer measuring unit 102, The jitter buffer measurement unit 103 collects the average packet arrival time variation averages and inputs them to the fuzzy inference engine unit 105.

  The user setting terminal 109 inputs various reception parameters (specifically, reception buffer amount and transmission buffer amount) set by the user to the fuzzy inference engine 105 via the user setting I / F 108 and the information input unit 107. In addition, the user setting terminal 109 inputs an instruction to modify (including addition, deletion, etc.) of the fuzzy inference rule set by the user to the fuzzy inference rule changing unit 106 via the user setting I / F 108. The fuzzy inference rule changing unit 106 corrects (including addition, deletion, etc.) fuzzy inference rules used by the fuzzy inference engine unit 105 in accordance with an instruction from the user setting terminal 109.

  FIG. 2 is a block diagram illustrating a configuration example of the fuzzy inference engine unit of FIG. The fuzzy inference engine unit 105 is configured, for example, as shown in FIG. 2, and includes information collected by the information input unit 107 (number of packets, increase / decrease amount of packets, variation average value of packet arrival times, user setting terminal 109 A predetermined multi-fuzzy inference rule (see FIG. 3 to be described later) is applied to the user-oriented items (specifically, to emphasize sound quality and delay avoidance) to execute multi-multistage fuzzy inference (see FIG. 4 to be described later). Based on the obtained fuzzy inference result, a change instruction is given to the changing means 110, 111, 112 and the packet transmission control unit 506.

  That is, the changing unit 110 of the packet reception control unit 502 changes the packet reception control operation of the packet reception control unit 502 to start, stop, or restart of reception, and the changing unit 111 of the reception buffer 503 sets the buffer amount of the reception buffer 503. The setting is changed, the changing unit 112 of the jitter buffer 505 changes the setting of the buffer amount of the jitter buffer 505, and the packet transmission control unit 506 changes the packet transmission control operation to start, stop, restart, etc. of transmission. As a result, the flow control of the reception buffer 503 and the control of the capacity of the jitter buffer 505 are dynamically associated and appropriately performed. As shown in FIG. 2, the fuzzy inference engine unit 105 includes fuzzy inference elements A201, B202, C203, D204, E205, and F206 arranged in multiple stages.

  FIG. 3 is a diagram showing a multiple fuzzy inference rule for each fuzzy inference element in FIG. The multiple fuzzy inference rules A301, B302, C303, D304, E305, and F306 shown in FIG. 3 are applied to these fuzzy inference elements. In this way, an inference form in which fuzzy inference elements defined by the multiple fuzzy inference rules are connected in multiple stages is called multiple multistage fuzzy inference.

The inference method uses the direct method, and in the present invention, the fuzzy relation synthesis rule that converts multiple rules into multiple fuzzy relations is used to suppress the spread of ambiguity from multiple fuzzy implications. Employs fuzzy implications by Godel having an effect. Let the antecedent and consequent fuzzy sets be B and C, and their membership functions μ _B (x) and μ _C (x), respectively. The fuzzy implication definition formula (Equation 1) is shown below.

  The fuzzy relation of the whole rule in the one-stage n-fold fuzzy inference is an intersection of n fuzzy relations. In other words, if the fuzzy relationship of the i-th rule is Ri and the multiple fuzzy relationship is MR,

Given in.

  The fuzzy relationship (Equation 2) of the multiple rules generated using the fuzzy implication (Equation 1) shown above is expressed as a two-dimensional fuzzy membership. The operation of multiple multistage fuzzy inference is shown. A known input is two-dimensionally synthesized by a multiple fuzzy relation and a t-norm operation, and the result is further projected to a one-dimensional result by a t-conorm operation. The inference result obtained in this way becomes an input to the next multiple rule. A typical t-norm and t-conorm operation is typically the min-max method using min for the t-norm operation and max for the t-conorm operation.

  Normally, multi-stage fuzzy inference, which performs defuzzy work such as taking the centroid of the fuzzy membership of the inference result every time fuzzy inference is performed, is the mainstream. In such fuzzy reasoning, the ambiguity generated in the reasoning process is negated, so that the meaning of the ambiguity of each reasoning may be erased. As described above, in the present invention, in multi-stage fuzzy inference using Rg fuzzy implication by Godel for synthesizing fuzzy relations, inference is performed while maintaining the ambiguity for each stage of inference without performing a fuzzy operation. It can be performed. Therefore, the information that the ambiguity of each inference acts more organically as the inference is repeated in multiple stages and propagates in the inference process can be kept intact. Passing ambiguity to the final conclusion while suppressing the spread of ambiguity can lead to inference results with more subtle nuances.

  This multi-stage fuzzy reasoning has the advantage that it can be ruled and modeled with linguistic connections for the relationship between different scales. For example, in the financial market, if the money supply rises, the amount of money circulating in the financial market increases, so the interest rate when a bank borrows funds from the BOJ decreases. Banks can raise funds at low interest rates, so they can lend to companies at low interest rates. Companies that have been able to borrow money at low interest rates can be modeled to increase capital investment and increase employment.

  FIG. 4 is a diagram for explaining a multi-stage fuzzy inference operation in the fuzzy inference engine unit 105 of FIG. In such a case, by superimposing a plurality of multiple fuzzy inference results, that is, as shown in FIG. 4, the first stage fuzzy inference is performed on a certain input 401 by the multiple fuzzy inference rule (1) 402, and the inference result is obtained. Is input to the next multi-fuzzy inference rule (2) 403 and the second-stage fuzzy inference is performed, so that the input of the previous inference result is input to the next multi-fuzzy inference rule (n ) By performing the multi-stage fuzzy inference to obtain the conclusion 405 by performing up to the n-th stage fuzzy inference according to 403, modeling can be realized as a multi-stage multi-stage fuzzy inference relatively easily. This multi-stage fuzzy inference is difficult to directly relate to, for example, from the aforementioned money supply to the increase in employment, and is advantageous in the case of obtaining an inference result whose scale is different from the input one.

  Also in this embodiment, in FIG. 4, the input 401 is a different measure such as the degree of packet congestion due to the amount of packet inflow as shown in FIG. 2, and the linguistic rule setting that emphasizes sound quality and delay avoidance as user settings. Multiple fuzzy inference rules (1) 402, multiple fuzzy inference rules (2) 403, which are multiple multi-stage fuzzy inferences that input the inference result of the previous stage to the next stage multi-fuzzy inference rule. , By using the multiple fuzzy inference rule (n) 404, the conclusion 405 of the control amount of the flow control and the jitter buffer can be obtained. The multi-stage fuzzy inference operation according to this embodiment will be specifically described below with reference to FIGS.

  A fuzzy inference operation in one stage of multiple multistage fuzzy inference is used as a fuzzy inference element. The fuzzy inference element is a fuzzy relation (Equation 2) using fuzzy implications (Equation 1) by Godel (Gedel) and fuzzy inference rules created by the direct method, and Min-Max as t-norm and t-conorm calculations It is shown as a black box of fuzzy inference operation that employs the method. The packet inflow amount 207 is input to the fuzzy inference element A201 as an input for the first stage inference. This packet inflow amount 207 is a difference value between the number of received packets measured by the received packet number measuring unit 101 and the number of transmitted packets measured by the transmitted packet number measuring unit 104. In the fuzzy inference element A201, any one of <rule 1-a> <rule 1-b> <rule 1-c> in the multiple fuzzy inference rule A301 shown in FIG. Apply. That is, if the packet inflow amount 207 is “increase”, “flow control is applied”, if “± 0”, “flow control is maintained”, and if “decrease”, “flow control is released”. And This is the content of the flow control amount (1) 208, which is the fuzzy inference result regarding the flow control obtained by the fuzzy inference element A201.

  The reception buffer packet amount 209 measured by the reception buffer measurement unit 102 is input to the fuzzy inference element B202 as an input of the first stage inference of the parallel fuzzy inference of the fuzzy inference element A201. In the fuzzy inference element B202, any one of <rule 2-a> <rule 2-b> in the multiple fuzzy inference rule B302 shown in FIG. That is, if the received buffer packet amount 209 is “large”, “flow control is applied”, and if it is “small”, “flow control is canceled”. This is the content of the flow control amount (2) 210 which is the fuzzy inference result regarding the flow control obtained by the fuzzy inference element B202.

  When inputting to the fuzzy inference element in the next stage, the inference result of the fuzzy inference element inferring in parallel with the same number of stages may be synthesized. The flow control amount (1) 208 obtained by the fuzzy inference element A201 and the flow control amount (2) 210 obtained by the fuzzy inference element B202 are synthesized by the flow control amount fuzzy synthesis processing unit 211. The flow control amount 212 that is the inference result is obtained. In the output as the fuzzy inference result, the defuzzy processing that eliminates the ambiguity of the output is not performed, and the fuzzy inference element, that is, the fuzzy inference element is input as it is. This becomes the input of the fuzzy inference element C203.

  The fuzzy inference element C203, which is the third stage of the multiple multistage fuzzy inference, is the first stage inference result, does not perform defuzzy, and directly uses this flow control amount 212 used as the second stage inference input by the direct method. Any one of <Rule 3-a> <Rule 3-b> <Rule 3-c> in the multiple fuzzy inference rule C303 shown in FIG. 3 is applied. That is, if the flow control amount 212 is “apply flow control”, “packet loss increase” is set, if “flow control is current status maintenance”, “packet loss is current status maintenance”, and “flow control is released” If there is, “packet loss reduction” is assumed. This is the content of the packet loss degree 213 that is the inference result obtained by the fuzzy inference element C203. In the output as the fuzzy inference result, the defuzzy processing that eliminates the ambiguity of the output is not performed, and the fuzzy inference element, that is, the fuzzy inference element is input as it is. Such a packet loss degree 213 is input to the fuzzy inference element D204.

  In the fuzzy inference element D204, which is the third stage of the multiple multistage fuzzy inference, this packet loss degree 213 is determined according to <rule 4-a> <rule 4-b in the multiple fuzzy inference rule D304 shown in FIG. > <Rule 4-c> is applied. That is, if the packet loss degree 213 is “packet loss increased”, “jitter buffer amount increased”, “no packet loss change”, “jitter buffer amount current maintenance”, and “packet loss decreased”. “Reduced jitter buffer amount”. This is the content of the jitter buffer control amount (1) 214 which is the fuzzy inference result regarding the jitter buffer control obtained by the fuzzy inference element D204.

  As parallel fuzzy inference performed with the same number of stages of the fuzzy inference element D204, the user-oriented items 215 such as sound quality emphasis and delay avoidance emphasis are input to the fuzzy inference element E205. In the fuzzy inference element E205, any one of <rule 5-a> <rule 5-b> in the multiple fuzzy inference rule E305 shown in FIG. That is, if the user-focused item 215 is “sound quality-focused”, “jitter buffer amount is increased”, and if “delay avoidance is focussed”, “jitter buffer amount is decreased”. This is the content of the jitter buffer control amount (2) 216, which is the fuzzy inference result regarding the jitter buffer control obtained by the fuzzy inference element E205.

  As parallel fuzzy inference performed at the same number of stages of the fuzzy inference element D204, the packet arrival variation average time 217 measured by the jitter buffer measuring unit 103 is input to the fuzzy inference element F206. In the fuzzy inference element F206, the average arrival time 217 of packet arrivals is changed to <Rule 6-a> <Rule 6-b> <Rule 6-c> in the multiple fuzzy inference rule F306 shown in FIG. Apply either. That is, if the packet arrival variation average time 217 is “packet variation increased”, “jitter buffer amount increased”, “no packet variation change”, “jitter buffer amount current”, and “packet variation decreased”. “Jitter buffer amount reduction”. This is the content of the jitter buffer control amount (3) 218 which is the fuzzy inference result regarding the jitter buffer control obtained by the fuzzy inference element F206.

  When deriving the final inference result of multiple multistage fuzzy inference, the inference results of the fuzzy inference elements D204, E205, and F206 that are inferring in parallel are synthesized in the third stage. Jitter buffer control amount (1) 214 obtained by fuzzy inference element D204, jitter buffer control amount (2) 216 obtained by fuzzy inference element E205, and jitter buffer control amount (3) 218 obtained by fuzzy inference element F206 The jitter buffer control amount fuzzy synthesis processing unit 219 normalizes each fuzzy inference result, synthesizes the fuzzy membership by the Min-Max method, etc., takes the center of gravity, and so on. A buffer control amount 220 is obtained.

  As described above, fuzzy threshold determination used when changing the control amount of the conventional flow control or changing the amount of jitter buffer, and a plurality of input flow control and control of the amount of jitter buffer are input. Since information with different scales can be comprehensively judged and performed in an optimal state, it is always appropriate to perform flow control accompanying frequent changes in the amount of received packets across thresholds, without frequent execution. The control amount (buffer amount) of the reception buffer and the jitter buffer can be dynamically associated and set.

  Further, the fuzzy inference engine unit 105 can input fuzzy inference obtained by fuzzy inference having a predetermined rule into fuzzy inference having further different rules, thereby enabling fuzzy inference with higher accuracy. Flow control and jitter buffer amount control.

  Since the fuzzy inference rule changing unit 106 is provided, the fuzzy inference element and the input condition are basically freely passed through the fuzzy inference rule changing unit 106 as long as the input conditions necessary for the output of the fuzzy inference engine unit 105 can be secured. Can be deleted. For example, if the user emphasis item 215 is not taken into consideration as a condition for deriving the jitter buffer control amount 220, the fuzzy inference element E205 and the user emphasis item 215 as an input can be deleted. In such a case, it is only necessary to secure an input necessary for deriving a conclusion, so the fuzzy inference element E205 can be deleted via the fuzzy inference rule changing unit 106.

  As described above, since the fuzzy inference rule changing unit 106 for changing the fuzzy inference rule of the fuzzy inference engine unit 105 is provided, it is easy to add and delete fuzzy inference elements, and the function of the fuzzy inference engine can be flexibly changed. In addition, more appropriate flow control and jitter buffer amount control can always be performed.

  As a result, the packet transmission control unit 506 performs transmission control in accordance with the flow control and jitter buffer control amount derived by the fuzzy inference engine unit 105 in conjunction with the packet reception control unit 502, so that the packet reception device 507 is controlled. Packets with more stable call quality can be implemented.

  As described above, according to the present invention, when packetized voice data is transmitted / received, the control amount of the reception buffer and the jitter buffer can be adaptively controlled using fuzzy logic, so that it is always optimal. A packet communication device and a packet communication method that can provide a good call quality to a packet receiving device can be obtained.

  As described above, the packet communication apparatus and the packet communication method according to the present invention are useful for performing transmission / reception of packetized voice data at an optimum call quality at all times.

1 is a block diagram of a packet communication apparatus according to an embodiment of the present invention. Block diagram showing a configuration example of the fuzzy inference engine unit of FIG. The figure which shows the multiple fuzzy inference rule of each fuzzy inference element of FIG. The figure explaining the multistage fuzzy inference operation | movement in the fuzzy inference engine part of FIG. Block diagram showing a configuration example of a conventional packet communication device The figure explaining the structure and control operation principle of the receiving buffer of FIG. The figure explaining the control operation principle of the jitter buffer of FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Packet communication apparatus 101 Reception packet number measurement part 102 Reception buffer measurement part 103 Jitter buffer measurement part 104 Transmission packet number measurement part 105 Fuzzy inference engine part 106 Fuzzy inference rule change part 107 Information input part 108 User setting interface (user setting I / F)
109 User Setting Terminal 110 Packet Reception Control Unit Changing Unit 111 Reception Buffer Changing Unit 112 Jitter Buffer Changing Unit 201 Fuzzy Inference Element A
202 Fuzzy inference element B
203 Fuzzy Inference Element C
204 Fuzzy inference element D
205 Fuzzy inference element E
206 Fuzzy inference element F
207 Packet inflow amount 208 Flow control amount (1)
209 Receive buffer packet amount 210 Flow control amount (2)
211 Flow control amount fuzzy synthesis processing unit 212 Flow control amount 213 Degree of packet loss 214 Jitter buffer control amount (1)
215 User-oriented items 216 Jitter buffer control amount (2)
217 Average packet arrival variation time 218 Jitter buffer control amount (3)
219 Jitter buffer control amount fuzzy synthesis processing unit 220 Jitter buffer control amount 301 Multiple fuzzy inference rule A (for fuzzy inference element A)
302 Multiple fuzzy inference rule B (for fuzzy inference element B)
303 Multiple Fuzzy Inference Rules C (for Fuzzy Inference Element C)
304 Multiple fuzzy inference rules D (for fuzzy inference element D)
305 Multiple Fuzzy Inference Rule E (for Fuzzy Inference Element E)
306 Multiple Fuzzy Inference Rule F (for Fuzzy Inference Element F)
401 Input to Multiple Multistage Fuzzy Reasoning 402 Multiple Fuzzy Reasoning Rule (1)
403 Multiple Fuzzy Inference Rules (2)
404 Multiple Fuzzy Inference Rules (n)
405 Result of n-stage multiple fuzzy inference 501 Network 502 Packet reception control unit 503 Reception buffer 505 Jitter buffer 504 Reception buffer amount measurement unit 506 Packet transmission control unit 507 Packet reception device

Claims (6)

A received packet number measuring means for measuring the number of packets received by a packet reception control means for controlling transmission and reception of packets with a transmission source;
A reception buffer measuring means for measuring the number of stored packets in a reception buffer for temporarily storing packets input from the packet reception control means;
A jitter buffer measuring means for measuring an arrival variation average time of packets arriving from the reception buffer to a jitter buffer for leveling delay fluctuations of packets input from the reception buffer;
A transmission packet number measuring means for measuring the number of packets transmitted by a packet transmission control means for controlling transmission of the stored packets of the jitter buffer to a receiver;
Measurement information collection means for collecting each measurement result of the received packet number measurement means, the reception buffer measurement means, the jitter buffer measurement means, and the transmission packet number measurement means;
A fuzzy inference engine unit that executes a fuzzy inference by applying a predetermined fuzzy inference rule to each measurement result collected by the measurement information collecting means;
Based on the fuzzy inference result of the fuzzy inference engine unit, the reception control operation of the packet reception control unit is changed, the settings of the buffer amounts of the reception buffer and the jitter buffer are changed, and the reception of the packet transmission control unit is performed. A packet communication apparatus comprising: changing means for changing an original transmission control operation. 2. The packet communication apparatus according to claim 1, further comprising fuzzy inference rule changing means for changing the fuzzy inference rule applied by the fuzzy inference engine unit. 2. The packet communication apparatus according to claim 1, wherein the fuzzy inference engine unit inputs a fuzzy inference result obtained by fuzzy inference having a predetermined rule to a fuzzy inference having further different rules. A first step of measuring the number of packets received by a packet reception control means for controlling transmission and reception of packets with a transmission source;
A second step of measuring the number of packets stored in a reception buffer for temporarily storing packets input from the packet reception control means;
A third step of measuring an arrival variation average time of packets arriving from the reception buffer in a jitter buffer that equalizes delay fluctuations of packets input from the reception buffer;
A fourth step of measuring the number of packets transmitted by packet transmission control means for controlling transmission of the stored packets of the jitter buffer to the receiver;
A fifth step of collecting measurement results obtained in each of the first to fourth steps;
A sixth step of executing fuzzy reasoning by applying a predetermined fuzzy reasoning rule to each measurement result collected in the fifth step in the fuzzy reasoning engine unit;
The reception control operation of the packet reception control unit is changed according to the fuzzy inference result of the fuzzy inference engine unit, the setting of each buffer amount of the reception buffer and the jitter buffer is changed, and the reception source of the packet transmission control unit is sent to the reception source A seventh step of changing the transmission control operation of
A packet communication method comprising: The packet communication method according to claim 4, further comprising an eighth step of changing the fuzzy inference rule applied by the fuzzy inference engine unit. 5. The packet communication method according to claim 4, wherein the fuzzy inference engine unit inputs a fuzzy inference result obtained by fuzzy inference having a predetermined rule into a fuzzy inference having further different rules.

Images