JP2003188731A - Variable rate encoding method, encoder and decoder - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To control the packet sending rate in response to the congestion of a network when performing communications using animation or audio on a packet network such as the Internet or Ethernet (R). <P>SOLUTION: An information source is encoded by an encoding means 11, code words encoded by the encoding means 11 are temporarily stored by a buffer 12, and the code words are outputted at an applied service rate μ. The service rate μ is determined by a service speed determining means 16 while using either a target value λ of the service speed or a code word quantity (k) stored in the buffer 12, and an encoding parameter (v) required for determining an encoding rate in the case of encoding the information source by the encoding means 11 is determined by an encoding parameter determining means 15 while using any one of the target value λ of the service rate, the code word quantity (k) stored in the buffer and the service rate μ. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

The present invention relates to a variable rate coding method, a coding device and a decoding device.

[0002]

2. Description of the Related Art There is known a variable rate coding method in which a coding rate is changed in accordance with an information source when coding an image or a sound. For example, the conventional encoding method used in MPEG2, MPEG4, etc. performs encoding at a predetermined fixed rate or a variable rate in which the encoding rate changes according to the information source. Although best-effort communication in packet networks such as the Internet can be used at low cost, the bandwidth and packet discard rate are not guaranteed. Also, in order to avoid network congestion, a congestion control function that adjusts the packet transmission rate is required.

[0003]

When performing communication using moving images and voice in a packet network such as the Internet or Ethernet, it is necessary to control the packet transmission rate according to the congestion of the network. However, MPEG2 and MPE
In the conventional coding method such as G4, the coding rate cannot be dynamically changed in response to the congestion of the packet network. Therefore, when these encoding methods are used, there is a problem in that the packet network becomes congested and information transfer is hindered.

The present invention has been made in consideration of the above circumstances, and when performing communication using moving images or voice in a packet network such as the Internet or Ethernet, the packet transmission speed is adjusted according to the network congestion. An object of the present invention is to provide a variable rate coding method, a coding device, and a decoding device that can be controlled.

[0005]

In order to achieve the above object, a variable rate coding method of the present invention includes a coding step of coding information from an information source, and a coding step performed by the coding step. The service speed is calculated by using at least one of an output step of temporarily storing a codeword in a buffer and outputting the codeword at a given service speed, and a target value of the service speed or the amount of codewords stored in the buffer. In the encoding step, the service rate determining step is performed, and at least one of the target value of the service rate, the codeword amount accumulated in the buffer, and the service rate is used. A coding parameter determining step for determining a coding parameter required to determine a coding speed when coding information. Is Dressings. In the encoding step, information from the information source is encoded at a plurality of different encoding rates, a main codeword that is a codeword encoded at the highest encoding rate, and a main codeword that is lower than the main codeword. The auxiliary codeword, which is a codeword obtained by encoding at the encoding speed, is output. Further, the information from the information source is at least one of audio information and image information, and the encoding parameter is an image size of an image obtained by down-sampling an input image, an image frame rate, or information of The number of information quantization bits that determines the quantization width, the difference frame rate that is the ratio of the number of frames that perform inter-frame predictive coding and the number of frames that perform intra-frame coding, the coding speed of audio information, and the speed of auxiliary codewords. At least one of an auxiliary code word rate, which is the speed ratio of the main code word, an auxiliary code word transmission delay time, which is the difference between the transmission times of the main code word and the auxiliary code word, and a transmission packet length. Is.

Furthermore, the encoding parameter determining step includes a part or all of the encoding parameters,
By referring to a table showing the relationship between the target value of the service rate, the codeword amount accumulated in the buffer, and at least one of the service rates, a part or all of the encoding parameters can be obtained. It is supposed to be decided. Furthermore, the relationship between the encoding parameter and at least one of the target value of the service rate, the codeword amount accumulated in the buffer, and the service rate, which are described in the table, is changed. The table quantization scale value for is designed so that it can be updated. Furthermore, at least one of the encoding parameters or the table quantization scale value can be changed from the receiving side. Furthermore, the service speed in the service speed determining step is changed by changing at least one of a delay time and a packet arrival status at the time of returning an ACK packet for transmitting the packet arrival status at the reception side to the transmission side. The operation of determining the target value of is adjusted.

Furthermore, the coding apparatus of the present invention is a coding means for coding information from an information source, and a code word coded by the coding means is temporarily stored and output at a given service speed. Buffer, a service rate determining means for determining the service rate using at least one of the target value of the service rate or the codeword amount accumulated in the buffer, the target value of the service rate, and the buffer Required to determine the coding rate at which the coding means codes the information from the information source using at least one of the accumulated codeword amount and the service rate. And a coding parameter determining means for determining a coding parameter.
Furthermore, the decoding device of the present invention is a main codeword, which is a codeword in which information from an information source transmitted on a communication network is encoded at the highest encoding speed, and an encoding lower than the main codeword. A decoding device that receives and decodes an auxiliary codeword that is a codeword encoded at a speed, and if the information cannot be reproduced using the main codeword, the information is output using the corresponding auxiliary codeword. It has a means for reproducing.

[0008]

BEST MODE FOR CARRYING OUT THE INVENTION The variable rate coding method of the present invention,
An encoding device to which the encoding method is applied, and a decoding device that decodes a codeword encoded by the encoding device and transmitted to a network, for example, compresses and encodes a signal input from a video camera or a microphone, and It is applied to a communication system that transmits data using the best effort communication of a packet network such as the Internet.

FIG. 1 is a block diagram showing the overall configuration of a communication system configured using an embodiment of an encoding device and a decoding device to which the variable rate encoding method of the present invention is applied. Here, the transmitting terminal 10 operates as the encoding device of the present invention, and the receiving terminal 20 operates as the decoding device of the present invention. In this figure, the signal input to the transmission terminal 10 from an information source such as the video camera 1 or the microphone 2 is
The encoding means 11 performs compression encoding based on the encoding parameter v. The code word output from the encoding means 11 is temporarily stored in the buffer 12 and output from the buffer 12 at the service speed μ. The congestion control means 14 determines the target value of the coding rate (target value of the service rate) λ by grasping the state of the network by receiving the ACK packet transmitted from the receiving side or the like. Since the coding speed depends not only on the coding parameter v but also on the input image, it cannot be uniquely determined only from the coding parameter v. Therefore,
Coding parameter v such that a desired coding speed is obtained
Is calculated by the coding parameter determination means 15 from the target value λ of the coding rate, the queue length (the amount of code words accumulated in the buffer 12) k, the service rate μ, and the like. Further, the service rate μ for outputting the codeword from the buffer 12 is determined by the service rate determining means 16 from the target value λ of the encoding rate and the queue length k, so that the codeword overflow in the buffer 12 is prevented. The code word output from the buffer 12 at the service rate μ is packetized by the packetizing means 13 and transmitted to the network 3 such as the Internet.

In the receiving terminal 20, the packet receiving means 21 receives the packet, and the decoding means 22 decodes the code word in the packet. The decoded signal is output to the display 4 and the speaker 5 after absorbing the fluctuation of the delay time by the delay fluctuation absorbing means 23. Further, when packet discard is detected by the packet discard detection means 24,
The signal corresponding to the discarded packet is interpolated by the packet interpolating means 25, and the display 4 and the speaker 5 are interpolated.
Is output to.

Here, in the encoding means 11, the information from the information source is encoded at a plurality of different encoding rates,
A codeword obtained by encoding at the highest encoding speed (referred to as "main codeword") and a codeword obtained by encoding at a lower encoding speed than the main codeword (referred to as "auxiliary codeword"). )
May be generated and stored in the buffer 12, the main codeword is transmitted first from the packetizing means 13, and the auxiliary codeword is transmitted with a predetermined delay time.
In this case, when the receiving terminal 20 (decoding device) cannot reproduce the information using the main codeword due to the reason that the packet of the main codeword has not arrived, the corresponding auxiliary It becomes possible to reproduce the information by using the code word.

FIG. 2 is a diagram showing a processing procedure of an embodiment of the variable rate coding method of the present invention executed in the transmission terminal 10. As shown in this figure, the encoding method according to the present embodiment includes steps S1 to S8. In step S1, the service rate μ, the target value of the service rate (the target value of the encoding rate) λ,
And the initial value of the coding parameter v is set. In step S2, information is input from the information source (video camera 1, microphone 2). Step S3 is a step executed by the encoding means 11, and encodes the information source according to the given encoding parameter v and outputs the code word. In step S4, the code word is stored in the buffer 12. Step S5 is a step executed by the service rate determining means 16, and based on at least one of the target value λ of the service rate and the codeword amount k stored in the buffer 12, the codeword from the buffer 12 is determined. The service speed μ that outputs is determined. At the beginning, the initial value set in step S1 is used. Step S6
Is a step executed by the congestion control means 14 and determines the target value λ of the service speed based on the network condition such as the reception condition of the ACK packet. Step S7 is a step executed by the encoding parameter determining means 15, and the encoding parameter is based on the target value λ of the service rate, the service rate μ, or the code word amount k accumulated in the buffer 12. Determine v. In step S8, the code word is output from the buffer 12 according to the service speed μ. Then, returning to the step S2, the steps S2 to S8 are repeatedly executed until the information transmission is completed.

The processing of the above steps will be described in detail below. In step S3, the information source is encoded according to the given encoding parameter v and a code word is output. Note that here, a code word (main code word) obtained by encoding the information from the information source at a plurality of different encoding rates and encoding at the highest encoding rate and an encoding lower than the main code word. It is assumed that a codeword (auxiliary codeword) obtained by encoding at speed is output. The encoding parameter v
Is the image size w of the image obtained by down-sampling the input image, the image frame rate f, the number of information quantization bits q that determines the quantization width of information, the number of frames for inter-frame predictive coding, and the intra-frame code. The differential frame rate r, which is the ratio of the number of frames to be encoded, the coding rate c of audio information, the auxiliary codeword rate a, which is the ratio of the speed of the auxiliary codeword and the speed of the main codeword, the main codeword and the auxiliary codeword It is described by at least one of the auxiliary codeword transmission delay time τ, which is the difference between the transmission times of, the transmission packet length ν, and the table quantization scale s. The table quantization scale s will be described later.

First, moving picture coding when the information source is a moving picture will be described with reference to the functional block diagram of the picture coding means shown in FIG. The moving image encoding method is an encoding method that uses the height of spatial correlation and temporal correlation of moving images, and includes a frame rate f, a differential frame rate r, an image size w, and the number of information quantization bits. This is a method in which q or the like can be changed as the encoding parameter v.

For the information input in step S2,
The image frame rate f is controlled by performing encoding only on the frame at the timing corresponding to the image frame rate f (the number of frames transmitted per unit time), and not encoding the other frames. Here, the image frame rate f is applied (changed) before the intra-frame coded frame is coded, and is constant until the next intra-frame coded frame. Also, the length of time for encoding one frame and inputting to the buffer is
The length of the frame interval is set according to the image frame rate f. Then, it is assumed that one intraframe coded frame and the number of interframe predictive coded frames based on the differential frame rate r are repeated. Here, the image size w, the auxiliary code word rate a, and the differential frame rate r are applied before the intra-frame coded frame is coded, and are constant until the next intra-frame coded frame.

For the intra-coded frame, the main codeword is determined by the following processing. First, the image to be encoded is down-sampled (31) to the image size w, and is further divided into 16 × 16 image areas. A discrete cosine transform coefficient is determined for each image region by the discrete cosine transform (33), and the coefficient is quantized (34) by the number of information quantization bits q. Finally, the main codeword is determined by variable length coding (35) for the quantization result. The auxiliary codeword is determined as follows. First, the image to be encoded is down-sampled (31) to the image size w ′ and divided into 16 × 16 image areas. The discrete cosine transform (33) is used to determine the discrete cosine transform coefficient for each image region, and the coefficient is quantized (34) with the information quantization bit number q ′. An auxiliary code word is determined for the quantization result by variable length coding (35). The image size w ′ and the information quantization bit number q ′ are determined from the auxiliary codeword rate a, the image size w and the information quantization bit number q. The number of information quantization bits q and q ′ when determining the main codeword and the auxiliary codeword are 16 ×
It is applied every time a fixed number of 16 image areas are encoded.

Intra-frame coding After coding a frame,
An image to be referred to in a subsequent inter-frame predictive coding frame is determined by the following process and stored in the frame memory 40. First, the quantization result of the discrete cosine transform coefficient for each image region obtained when determining the main codeword is inversely quantized (36) with the information quantization bit number q,
Further, inverse discrete cosine transform (37) is performed. After this processing is performed on the entire image area, up-sampling (39) is performed to the size of the image to be encoded and stored in the frame memory 40.

For the inter-frame predictive coded frame, the main codeword is determined by the following processing. First, the encoding target image and the reference image stored in the frame memory 40 are down-sampled (41) to the image size w,
It is divided into 16 × 16 image areas. An error between the encoding target image and the reference image is obtained for each image region (32), and a discrete cosine transform coefficient is determined for this error by the discrete cosine transform (33). Further, the main codeword is determined by variable length coding (35) with respect to the result of quantizing (34) this coefficient with the information quantization bit number q. The supplementary code word is determined as follows. First, the image to be encoded and the reference image stored in the frame memory are down-sampled (31, 41) to an image size w ′,
It is divided into 16 × 16 image areas. An error between the encoding target image and the reference image is obtained for each image region (32), and a discrete cosine transform coefficient is determined for this error by the discrete cosine transform (33). Further, with respect to the result of quantizing (34) this coefficient with the information quantization bit number q ', an auxiliary code word is determined by variable length coding (35). The information quantization bit numbers q and q'when determining the main codeword and the auxiliary codeword are applied every time a fixed number of 16x16 image areas are coded.

After the inter-frame predictive coding frame is encoded, the image to be referred to in the subsequent inter-frame predictive coding frame is determined by the following processing and stored in the frame memory 40. First, the quantization result of the discrete cosine transform coefficient (3
4), inverse quantization (36) is performed with the information quantization bit number q, and further inverse discrete cosine transform (37) is performed. After this processing is performed on the entire image area, upsampling (39) is performed to the size of the image to be encoded, and the sum with the image stored in the frame memory (40) is calculated (38) and stored in the frame memory 40. Store.

Next, the encoding of audio information will be described with reference to the functional block diagram of the audio encoding means shown in FIG. The audio information input in the step S2 is, for example, 1 channel of 2 channels and 8 kHz sampling.
It is assumed to be 6-bit PCM. In audio encoding, the following two types of encoding methods are switched according to the audio encoding rate c. Further, the application (change) of the audio coding rate c is made every fixed time. When the audio coding rate c is larger than a specific value, quantization (51, 52) corresponding to the audio coding rate c is performed, and a code word is output with a fixed-length code corresponding to the quantization. The amount of code words at this time can be controlled in stages. When the audio coding rate c is equal to or lower than a specific value, coding (51, 53) is performed by a CELP (Code Exited Linear Prediction) method such as ITU G.729. The amount of code words at this time depends on the system. The above is the encoding process of step S3 in FIG.

Next, in step S4, the code words (main code word and auxiliary code word) encoded in this way are accumulated in the buffer 12.

In step S5, the service speed of the buffer
Determine μ. At that time, discarding the codeword in the buffer 12
In order to prevent this, it is necessary to successively update μ. Only
However, the arrival rate of code words to the buffer fluctuates and
Since it is difficult to measure, the relationship between queue length k and service speed μ
Is not uniquely determined. Therefore, in the present invention,
Apply a non-linear system control method. This control method is
i_maxAnd m_{i; max}Is a constant, i ∈ {1, 2, ..., i_max}, M ∈
{1,2, ..., m_{i; max}}, X_{i; m}System input, y_{i; m}To
System output, z_{i; m}= G_{zi; m}(x_{i; m}, y_{i; m})
Number and z_{i; m} ^*Z_{i; m}X is the target value of_{i; m}, Y
_{i; m}, And g_{zi; m}(x_{i; m}, y_{i; m}), Respectively dy_{i; m}/ Dx_{i; m}
≧ 0, ∂g_{zi; m}(x_{i; m}, y_{i; m}) / ∂x_{i; m}≧ 0, and ∂g
_{zi; m}(x_{i; m}, y_{i; m}) / ∂y_{i; m}Choose so that ≧ 0, and z
_{i; m}And z_{i; m} ^*X as small as possible_{i; m}
Decide. Also, c_{i; m}Is a variable that represents the characteristics of the system,
σ_{ci; m}C_{i; m}, A variable that represents the fluctuation of_{i; m}X_{i; m}change of
Parameter that determines speed, D_{i; m}When (x, y) is x = y
D_{i; m}D when (x, y) = 1 and x ≠ y_{i; m}(x, y)> 1
If it is a function, x_{i; m + 1}= A_{i; m}And y_{i; m + 1}= Σ_{ci; m}
/ D_{i; m}(z_{i; m} ^*, z_{i; m}), The system
Adjust the parameters that determine the rate of change.

Calculate x _{i; m} as follows _: In a discrete time system, z _{i; m} (t) is defined by the following equation.

[Equation 1] Here, t represents a discrete time and d represents a delay time. this,

[Equation 2] In is approximated, the unknown coefficients zeta _i; estimate zeta ^ _i of _{m; m} is obtained by the following equation.

[Equation 3] Here, ω ₀ and ω ₁ are real numbers such that ω ₀ + ω ₁ = 1 and ω ₀ and ω ₁ > 0, W [·] is a moving average, and ε _{i; m} (t) is a random variable with an average of 0. .

The minimum variance control (KJ Astrom, "In
introduction to stochastic controltheory, "Academic
Press, Inc., London, 1970), z _{i; m} (t)
Is obtained by the following equation.

[Equation 4] Here, assuming that q _{i; m} is sufficiently large and the change in x _{i; m} (t) is sufficiently small,

[Equation 5] Holds. x _{i; m + 1} (t-1)

[Equation 6] Then, substituting equation (6) and equation (7) into equation (5),

[Equation 7] Is obtained. The formula (8) calculates x _{i; m} (t).

X _{i; m} (t) is the control input of the system, y _{i; m} (t)
Is the control output of the system, x _{i; m + 1} in equation (8)
(t) is a parameter that determines the rate of change of x _{i; m} (t).
For example, if D _{i; m} (x, y) is | x−y | +1,

[Equation 8] Becomes By replacing m in equation (8) with m + 1,
x _{i; m + 1} (t) that gives the desired property for the variation of x _{i; m} (t)
(A parameter that determines the changing speed of x _{i; m} (t)) can be determined.

In the present embodiment, R _μ = (λ-c) / μ
And Then, the above-mentioned nonlinear system control method is applied to the control of R _μ . _Assuming i _max = 1 and m _{i; max} = 2,
For each variable in (8), x _{i; m} (t) = R _μ (t), y _{i; m} (t) = k
_{(t), x i; m} + 1 (t) = p R ( constant), the target value of the ^{_{R μC R μ, z * =}} z R * = R μC bR k C, z R (t) = R μ (td) ^bR k
By substituting (t), the following equation is obtained.

[Equation 9] Here, d is a control delay. Therefore,

[Equation 10] Therefore, an appropriate service speed μ (t) can be determined. The average value kbar of k and the average value Rbar _μ of R _μ obtained as a result of the above control are the intersections of the characteristic curve and the control curve, as shown in FIG.

Other methods can also be adopted.
In this case, the service speed μ is determined by the following equation.

[Equation 11] Figure 6 shows the queue length k and service speed μ according to the above equation (12).
Shows the relationship. Here, K is the total amount of code words that can be stored in the buffer 12, k is the amount of code words in the buffer (queue length), and α, β, and γ are given constants. When the service speed μ is determined by the above equation, a constant α that does not cause the discard of the codeword in the buffer 12 is determined by the following. First,

[Equation 12] Substituting into equation (12),

[Equation 13] Is obtained. Here, 0 ≦ γ ≦ 1.

The relationship among the parameters α, β, γ, λ and the target value k _C of the code word amount is

[Equation 14] Can be obtained at. Transform this,

[Equation 15] Then, α is obtained by the following equation.

[Equation 16] Therefore, substituting equation (14) into equation (13),

[Equation 17] Determines the service speed μ. The service speed μ may be determined by any of the above methods.

In step S6, for example, MAQS (H.S.
atoh, "TCP flow control for maximum throughput and
QoS guarantee, "Internet Workshop '99, Japan, pp.
280-285, Feb. 1999, H. Satoh, M. Ishiba, T. Kobay
ashi, T. Enomoto, H. Morikawa, and K. Shimamura, "
Experimental Evaluation of Maximum Throughput and
QoS guaranteed Flow Control over Japan Gigabit Net
work, "IEICE Tech. Rep. IN200-211, pp. 217-223, Ma
y, 2001.) A packet transmission rate determined by a congestion control method such as that shown in FIG. For example, using the system control method described above, i _max = 1, m _{i; max}
= 2 and applies to congestion control of TCP or RTP. Here, λ (n) is the transmission rate of the packet n, r (n) is the probability of packet discard (packet discard rate), and n _ACK is the latest packet number that received ACK.

First, the case of m = 1 will be described. λ
(n) is the system input x ₁ , r (n) is the system output y ₁ , φ _{1 *}
Target value of control z ₁ ^*, φ ₁ and (n) as a state variable z _1, phi ₁
^* And φ ₁ (n) are defined by the following equations.

[Equation 18] Here, ν is the packet length and RTT is the round trip delay. Also,

[Formula 19] May be

Z _{i; m} ^* , z _{i; m} (t), x in the above equation (8)
_{i; m} (t), x _{i; m} (td) and x _{i; m + 1} (t-1) are represented by φ ₁ ^* , φ ₁ (n
_ACK ), λ (n), λ (n _ACK ) and a ₂ (n),

[Equation 20] Is obtained.

Next, with m = 2, a ₂ (n) is derived. σ _λ
(n) is defined as the system variation σ _c1 by the following equation.

[Equation 21] The state variable φ ₂ (n) is defined by the following equation.

[Equation 22]

Z _{i; m} ^* , z _{i; m} (t), x in the above equation (8)
_{i; m} (t), x _{i; m} (td), and x _{i; m + 1} (t-1) are φ ₂ ^* , φ
_{Substituting 2} (n _ACK ), a ₂ (n), a ₂ (n _ACK ), and the constant a ₃ ,

[Equation 23] Is obtained.

As described above, the calculation of the packet transmission rate by the congestion control means 14 is transmitted from the receiving side A
Since it is based on the reception status of the CK packet, the packet transmission speed (service speed) by the congestion control means 14 is changed by changing the transmission timing (transmission delay time) and the reception status of the ACK packet transmitted from the reception side to the transmission side. It is possible to control the algorithm for calculating the target value λ) of. For example, ACK at the receiving terminal 20
It is possible to control the congestion control algorithm by controlling the reception status of the packet or delaying the time of transmitting the ACK packet corresponding to the arrival packet by a predetermined time (ACK packet delay), and controlling this reception parameter. it can. This is the receiving terminal 20
May independently set the reception parameter, or the transmission parameter may be transmitted from the transmission terminal 10 to the reception terminal 20 and the operation of the reception terminal 20 may be controlled. Further, the number of transmission packets corresponding to one ACK packet may be changed at any time.

Next, in step S7, the step S6 is executed.
The coding parameter v is determined on the basis of the target value λ of the service speed determined in. In this example, v = (w, f, r,
a, q, c), that is, the image size w, the image frame rate f, the difference frame rate r, the auxiliary code word rate a, the information quantization bit number q, and the encoding rate c of the audio information that can be changed. Shall be selected as. First, the coding rate c of audio information is determined according to the target value λ of the service rate. Next, according to the table shown in FIG. 7, w, f,
r and a are determined from the target value λ of the service speed. Here, λ ₀ , λ ₁ , ... Are constants, i is a table index, and s is a table quantization scale. By changing the table quantization scale s, the range of the value of (λ-c) in each row of the table shown in FIG. 7 is changed, and the larger s, the higher the coding rate. Since the relationship between the coding rate and the number of information quantization bits q varies depending on the input image, the optimum coding parameter v for the given target value λ of the service rate also varies. Therefore, the table quantization scale s is sequentially updated to maintain the relationship between the target value λ of the service speed and the optimum coding parameter v.

Since the table quantization scale s and the queue length k have a relationship of dk / ds> 0, the characteristic curve of FIG. 8 is obtained, but this characteristic curve is an unknown time-varying nonlinear function. Therefore, the above-mentioned control method of the nonlinear system is applied to the control of the table quantization scale s. That is, assuming that i _max = 1 and m _{i; max} = 2, the variables of the formula (8) are represented by x _{i; m} (t) = s (t), y _{i; m} (t) = k (t), x _{i; m + 1} (t) ＝
p _s (constant), s _C is the target value of the table quantization scale s, k _C is the target value of the queue length k, z ^* = z _s ^* = s _C ^bs k _C ,
By substituting z _s (t) = s (td) ^bs k (t), the following equation is obtained. Here, d is a control delay.

[Equation 24] This makes it possible to determine an appropriate table quantization scale s. Average value k of k obtained as a result of the above control
The average value sbar of bar and s is the intersection of the characteristic curve and the control curve, as shown in FIG.

Further, as in the case of the table quantization scale s, the information quantization bit number q and the queue length k have a relationship as shown in FIG. However, since the relationship between the coding rate and the information quantization bit number q varies depending on the input image, the information quantization bit number q for a given target value λ of the service rate cannot be uniquely obtained. Therefore, applying equation (8) as in equation (24),

[Equation 25] Can determine the number q of information quantization bits. Where q
_C is a target value of the information quantization bit number q, and z _q (t) = q (td) ^{bq
_{k (t), z q *}} = q C bq k C, p q and b _q are constants, d is a control delay. The average value kbar of k and the average value qbar of q obtained as a result of the above control are the intersections of the characteristic curve and the control curve of FIG. This makes it possible to adaptively control the number q of information quantization bits corresponding to the given target value λ of the service speed.

In the above embodiment, the coding parameter v or the information quantization bit number s is changed in the transmitting terminal 10, but the receiving terminal 20 transmits the control information to the transmitting terminal 10. Then, the encoding parameter v or the information quantization bit number s can be adjusted. In this case, for example, in a remote monitoring system or the like, it is possible to select an encoding method according to the characteristics of information required on the receiving side.

In the above description, the coding parameter v corresponding to the target value λ of the service speed is used as a reference, but the service speed μ is determined.
Alternatively, the value of the code word amount k accumulated in the buffer may be used as a reference to determine the corresponding coding parameter v. Alternatively, a plurality of the target value λ of the service rate, the service rate μ, or the code word amount k accumulated in the buffer may be used as a reference. Further, in the above table, the example in which the image size w, the image frame rate f, the differential frame rate r and the auxiliary code word rate a are selected as the encoding parameter v is shown, but the present invention is not limited to this. The auxiliary codeword transmission delay time τ, the transmission packet length ν, and the like, which are the differences between the transmission times of the main codeword and the auxiliary codeword, may be determined.

[0040]

As described above, according to the variable rate coding method, the coding device, and the decoding device of the present invention, when performing communication using a moving image or voice in a packet network such as the Internet or Ethernet, It is possible to control the packet transmission rate according to the network congestion. Further, this makes it possible to avoid congestion of the packet network that occurs when a conventional variable rate coding method such as the conventional MPEG2 or MPEG4 is used, so that voice or image communication can be performed without putting pressure on other traffic. .

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a communication system having an encoding device and a decoding device to which the variable rate encoding method of the present invention is applied.

FIG. 2 is a flowchart showing a processing procedure in an embodiment of a variable rate coding method of the present invention.

FIG. 3 is a functional block diagram for explaining image encoding means.

FIG. 4 is a functional block diagram for explaining a voice encoding unit.

FIG. 5 is a diagram showing a characteristic curve and a control curve of Rμ and queue length k.

FIG. 6 is a diagram showing a control curve of the service speed μ in another method of calculating the service speed μ.

FIG. 7 is a diagram showing an example of a quantization table.

FIG. 8 is a diagram showing a characteristic curve and a control curve of a table quantization scale s and a queue length k.

FIG. 9 is a diagram showing a characteristic curve and a control curve of an information quantization bit number q and a queue length k.

[Explanation of symbols]

1 video camera 2 microphone 3 network 4 display 5 speakers 10 Sending terminal 11 Encoding means 12 buffers 13 Packetization means 14 Congestion control means 15 Encoding parameter determining means 16 Service speed determination means 20 Receiving terminal 21 packet receiving means 22 Decoding means 23 Delayed fluctuation absorbing means 24 Packet discard detection means 25 Packet discard interpolation means

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Hitoki Sato             Kanagawa Prefecture Yokosuka City Hikarinooka No. 3-4 Stock Association             YRP Mobile Communication Infrastructure Research Institute             Inside the laboratory (72) Inventor Toshio Kasama             2-3 Kandanishikicho, Chiyoda-ku, Tokyo Stock             Inside the Fuji Research Institute (72) Inventor Yoshimasa Yasukawa             2-3 Kandanishikicho, Chiyoda-ku, Tokyo Stock             Inside the Fuji Research Institute (72) Inventor Taiko Arima             2-3 Kandanishikicho, Chiyoda-ku, Tokyo Stock             Inside the Fuji Research Institute F term (reference) 5C059 KK34 LB01 LB05 MA00 MA04                       MA05 MA23 MA31 MC14 MC38                       ME01 ME17 PP04 RA08 RB03                       RB16 RC32 SS08 TA47 TA71                       TC20 TC25 TC27 TC37 TC45                       TD16 TD17 UA02 UA05 UA32                       UA33                 5C064 AA06 AB04 AC04 AC06 AC12                       AC16 AD02 AD06 AD14 AD16                 5D045 DA20                 5J064 AA00 BA16 BB03 BB10 BC01                       BC15 BC16 BC29 BD02

Claims (9)

[Claims] 1. A coding step of coding information from an information source, an output step of temporarily storing a codeword coded by the coding step in a buffer and outputting the codeword at a given service speed, and a service. A service speed determining step of determining the service speed using at least one of a target speed value and an amount of code words accumulated in the buffer; a target speed speed value, accumulated in the buffer; At least one of the code word amount and the service rate is used to determine the coding parameter required for determining the coding rate at the time of coding the information from the information source in the coding step. A variable rate coding method, comprising: a coding parameter determining step of determining. 2. The encoding step encodes information from the information source at a plurality of different encoding rates, a main codeword which is a codeword encoded at the highest encoding rate, and the main code. The variable rate coding method according to claim 1, wherein an auxiliary code word, which is a code word obtained by coding at a coding rate lower than that of the word, is output. 3. The information from the information source is at least one of audio information and image information, and the encoding parameter is an image size and an image frame rate of an image obtained by downsampling an input image. , The number of information quantization bits that determines the quantization width of information, the difference frame rate that is the ratio of the number of frames that perform interframe predictive coding and the number of frames that perform intraframe encoding,
Audio information coding rate, auxiliary codeword rate, which is the ratio of auxiliary codeword speed to main codeword speed, auxiliary codeword transmission delay time, which is the difference between the transmission time of the main codeword and auxiliary codeword, and the transmission packet The variable rate coding method according to claim 1 or 2, wherein the variable rate coding is at least one of the lengths. 4. The encoding parameter determination step comprises:
By referring to a table showing a relationship between some or all of the encoding parameters and at least one of the target value of the service rate, the amount of code words accumulated in the buffer, and the service rate 4. The variable rate coding method according to claim 3, wherein a part or all of the coding parameters are determined. 5. The relationship between the coding parameter and at least one of the target value of the service rate, the codeword amount accumulated in the buffer, and the service rate, which are listed in the table, 5. The variable rate coding method according to claim 4, wherein the table quantization scale value for changing can be updated. 6. The method according to claim 1, wherein at least one of the encoding parameters or the table quantization scale value can be changed by a receiving side. Item 6. The variable rate coding method according to any one of Items 5. 7. The service speed determining step according to claim 1, wherein at least one of a delay time when returning an ACK packet for notifying the transmitting side of the packet arrival status at the receiving side and a packet arrival status is changed. 7. The variable rate coding method according to claim 1, wherein the operation for determining the target value of the service rate can be adjusted. 8. An encoding means for encoding information from an information source, a buffer for temporarily accumulating codewords encoded by the encoding means and outputting at a given service speed, and a target value of the service speed. Alternatively, a service rate determining means for determining the service rate using at least one of the codeword amounts stored in the buffer, a target value of the service rate, the codeword amount stored in the buffer, and the A coding parameter for determining a coding parameter required to determine a coding rate when the coding means codes information from an information source by using at least one of service rates. An encoding device comprising: determining means. 9. A main codeword, which is a codeword coded at the highest coding rate, and information from an information source transmitted on a communication network is coded at a coding rate lower than the main codeword. A decoding device that receives and decodes an auxiliary codeword that is a codeword, and when the information cannot be reproduced using the main codeword, a unit that reproduces the information using the corresponding auxiliary codeword is provided. A decoding device having.