WO2011158421A1 - Dispositif de spécification de modèle - Google Patents

Dispositif de spécification de modèle Download PDF

Info

Publication number
WO2011158421A1
WO2011158421A1 PCT/JP2011/002304 JP2011002304W WO2011158421A1 WO 2011158421 A1 WO2011158421 A1 WO 2011158421A1 JP 2011002304 W JP2011002304 W JP 2011002304W WO 2011158421 A1 WO2011158421 A1 WO 2011158421A1
Authority
WO
WIPO (PCT)
Prior art keywords
model
packet
behavior
mathematical
behavior information
Prior art date
Application number
PCT/JP2011/002304
Other languages
English (en)
Japanese (ja)
Inventor
耕介 野上
Original Assignee
日本電気株式会社
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 日本電気株式会社 filed Critical 日本電気株式会社
Priority to JP2012520251A priority Critical patent/JPWO2011158421A1/ja
Publication of WO2011158421A1 publication Critical patent/WO2011158421A1/fr

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network

Definitions

  • the present invention relates to a model specifying device for specifying a mathematical model for estimating packet behavior.
  • a known service for transmitting / receiving content data representing content (video and / or audio) via a communication network (for example, an IP (Internet Protocol) network, a mobile communication network, and a wireless communication network).
  • a communication network for example, an IP (Internet Protocol) network, a mobile communication network, and a wireless communication network.
  • this type of service is, for example, IP phone, video conference, and video distribution such as movie or live.
  • a model specifying device that specifies a mathematical model for estimating the behavior of a packet transmitted over a communication network and estimates the behavior of the packet based on the identified mathematical model.
  • the model specifying device described in Patent Document 1 acquires behavior information indicating the behavior of a packet, and sets model parameters for specifying a mathematical model based on the acquired behavior information. presume. Then, the model specifying device estimates the behavior of the packet based on the mathematical model specified by the estimated model parameter.
  • the model specifying apparatus described in Non-Patent Document 1 uses an ARMA (Autogressive Moving Average) model as a mathematical model.
  • the ARMA model is expressed as shown in Equation 1.
  • y t and y ti are target time-series data (that is, here, behavior information at each of a plurality of time points).
  • p is the order of the part corresponding to the autoregressive model
  • q is the order of the part corresponding to the moving average model.
  • a i is a coefficient in the autoregressive model
  • b i is a coefficient in the moving average model.
  • e t and e ti are error terms.
  • Non-Patent Document 2 uses a hidden Markov model as a mathematical model.
  • Non-Patent Document 3 it is pointed out in Non-Patent Document 3 that when the network state changes between a congested state and a non-congested state, the behavior of the packet changes relatively greatly.
  • the number of model parameters of the mathematical model that can estimate the packet behavior with high accuracy often differs depending on the network state.
  • the number of model parameters corresponds to the number of internal states that is the number of states.
  • the number of model parameters corresponds to the order (that is, p and q in Equation 1).
  • the model specifying device is not configured to change the number of model parameters even when the network state changes. For this reason, there is a possibility that the model specifying device cannot specify a mathematical model capable of estimating the actual packet behavior with high accuracy.
  • the object of the present invention is to specify a model that can solve the above-mentioned problem that “the mathematical model that can estimate the actual packet behavior with high accuracy cannot be specified”. To provide an apparatus.
  • a model specifying device includes: Behavior information acquisition means for acquiring behavior information representing the behavior of a packet transmitted over a communication network; For each of a plurality of mathematical models specified by model parameters and for estimating the behavior of the packet, the number of the model parameters being different, the acquired behavior information A model parameter estimating means for estimating a model parameter for identifying the mathematical model, For each of the plurality of mathematical models, a degree of fitness representing the degree to which the behavior of the packet estimated by the mathematical model matches the behavior of the packet represented by the acquired behavior information is calculated. A fitness calculation means for Model selection means for selecting one mathematical model from the plurality of mathematical models based on the calculated fitness; Is provided.
  • a model specifying method includes: Obtain behavior information that represents the behavior of packets transmitted over the communication network, For each of a plurality of mathematical models specified by model parameters and for estimating the behavior of the packet, the number of the model parameters being different, the acquired behavior information Based on the model parameters for identifying the mathematical model, For each of the plurality of mathematical models, a degree of fitness representing the degree to which the behavior of the packet estimated by the mathematical model matches the behavior of the packet represented by the acquired behavior information is calculated. And In this method, one mathematical model is selected from the plurality of mathematical models based on the calculated fitness.
  • the model specifying program which is another embodiment of the present invention,
  • Behavior information acquisition means for acquiring behavior information representing the behavior of a packet transmitted over a communication network; For each of a plurality of mathematical models specified by model parameters and for estimating the behavior of the packet, the number of the model parameters being different, the acquired behavior information
  • a model parameter estimating means for estimating a model parameter for identifying the mathematical model, For each of the plurality of mathematical models, a degree of fitness representing the degree to which the behavior of the packet estimated by the mathematical model matches the behavior of the packet represented by the acquired behavior information is calculated.
  • a fitness calculation means for Model selection means for selecting one mathematical model from the plurality of mathematical models based on the calculated fitness; It is a program for realizing.
  • the present invention is configured as described above, and can specify a mathematical model that can estimate the actual packet behavior with high accuracy.
  • the communication device (model specifying device) 100 is connected to a communication line constituting a communication network (in this example, an IP (Internet Protocol) network).
  • the communication device 100 transmits / receives a packet to / from another communication device connected via a communication line.
  • the communication device 100 includes a central processing unit (CPU; Central Processing Unit) and a storage device (memory) (not shown).
  • the communication device 100 is configured to realize functions to be described later when a CPU executes a program stored in a storage device.
  • FIG. 1 is a block diagram showing functions of the communication apparatus 100 configured as described above.
  • the function of the communication device 100 includes a network state estimation unit 100A and a distribution control unit 100B.
  • the network state estimation unit 100A estimates model parameters for specifying a mathematical model for each of a plurality of network models (mathematical models) based on behavior information representing the behavior of packets transmitted over a communication network. To do. Furthermore, the network state estimation unit 100A calculates the fitness for each of the plurality of mathematical models, and selects a mathematical model that can estimate the behavior of the packet with the highest accuracy based on the calculated fitness.
  • the distribution control unit 100B determines a control parameter based on the mathematical model selected by the network state estimation unit 100A.
  • the control parameters are the redundancy that the packet has in order to perform error correction processing, and the encoding rate used when encoding the content represented by the packet.
  • the network state estimation unit 100A includes a behavior information acquisition unit (behavior information acquisition unit) 101, a model learning unit (model parameter estimation unit) 102, a fitness level calculation unit (a fitness level calculation unit) 103, and a fitness level estimation unit ( (Fitness estimation means) 104 and a model selection unit (model selection means) 105.
  • the behavior information acquisition unit 101 acquires behavior information indicating the behavior of a packet transmitted on the communication network.
  • the behavior information is information indicating a delay time that is a time required for the packet to reach the transmission destination from the transmission source, and the presence or absence of occurrence of a packet loss in which the packet does not reach the transmission destination.
  • the behavior information acquisition unit 101 acquires behavior information indicating the behavior of the packet received by the communication apparatus 100 within the immediately preceding acquisition cycle every time a preset acquisition cycle elapses.
  • the behavior information acquisition unit 101 may be configured to acquire behavior information representing the behavior of a packet received by a communication device other than the communication device 100 from the communication device.
  • the model learning unit 102 estimates a model parameter for specifying the mathematical model for each of the plurality of mathematical models based on the behavior information acquired by the behavior information acquisition unit 101.
  • the plurality of mathematical models are mathematical models having different numbers of model parameters.
  • Each mathematical model is a mathematical model for estimating packet behavior.
  • the model learning unit 102 estimates model parameters by performing a learning process.
  • the mathematical model is a hidden Markov model. Therefore, the number of model parameters changes according to the number of internal states, which is the number of states in the hidden Markov model.
  • the model parameters include a state transition probability, a symbol output probability, and an initial state probability.
  • the state transition probability is the probability that the state in the hidden Markov model will transition.
  • the symbol output probability is a probability distribution with the symbol value output when the state in the hidden Markov model transitions as a random variable.
  • the initial state probability is a probability distribution having the initial state in the hidden Markov model as a random variable.
  • the mathematical model may be a time series model (such as an autoregressive moving average model, an autoregressive model, or a moving average model).
  • the number of model parameters is the order in the time series model.
  • the fitness level calculation unit 103 calculates the fitness level for each of a plurality of mathematical models. As for the fitness, the behavior of the packet estimated by the mathematical model specified by the model parameter estimated by the model learning unit 102 matches the behavior of the packet represented by the behavior information acquired by the behavior information acquisition unit 101. Represents the degree to which
  • the goodness-of-fit estimation unit 104 determines the future based on the time series data composed of the goodness of fit calculated by the goodness-of-fit calculation unit 103 at each of a plurality of times before the current time. Estimate the goodness of time.
  • the fitness level estimation unit 104 estimates the fitness level in the future based on a time series model (such as an autoregressive moving average model, an autoregressive model, or a moving average model).
  • the model selection unit 105 selects one mathematical model from a plurality of mathematical models based on the fitness at a future time point estimated by the fitness estimation unit 104. In this example, the model selection unit 105 selects a mathematical model having the highest degree of fitness at a future time.
  • the distribution control unit 100B includes a feature information extraction unit (statistics calculation unit) 106 and a control parameter determination unit (reproduction quality estimation unit and control parameter determination unit) 107.
  • the feature information extraction unit 106 extracts (acquires) feature information based on the mathematical model selected by the model selection unit 105.
  • the feature information extraction unit 106 extracts feature information by performing random sampling based on the mathematical model selected by the model selection unit 105.
  • Feature information is information representing the quality of the communication network.
  • the feature information represents a statistical amount of packet behavior (for example, an average value of delay times or an occurrence rate of packet loss).
  • the feature information may represent the probability that a symbol string composed of a plurality of symbol values appears.
  • the symbol value is a value representing the behavior of the packet (for example, the magnitude of the delay time or the presence or absence of occurrence of packet loss).
  • the control parameter determination unit 107 estimates the reproduction quality of the content based on the packet based on the feature information extracted by the feature information extraction unit 106. Furthermore, the control parameter determination unit 107 determines a control parameter used when transmitting a packet based on the estimated reproduction quality of the content. At this time, the control parameter determination unit 107 determines the control parameter so as to maintain or improve the reproduction quality of the content.
  • the operation of the communication apparatus 100 described above includes an operation of selecting a mathematical model as shown in FIG. 2 and an operation of determining a control parameter as shown in FIG. First, the operation of selecting a mathematical model will be described with reference to FIG.
  • the communication device 100 receives a packet transmitted over a communication network.
  • the behavior information acquisition unit 101 acquires behavior information based on the received packet (step A01). For example, the behavior information acquisition unit 101 represents the delay time based on information included in the packet that represents the date and time when the packet was transmitted (transmission date and time) and information that represents the date and time when the packet was received (reception date and time). Get behavior information. Further, the behavior information acquisition unit 101 acquires behavior information indicating whether or not a packet loss has occurred by determining whether or not the sequence numbers included in the received packets are consecutive.
  • the model learning unit 102 estimates a model parameter for specifying the mathematical model by performing a learning process on each of the plurality of mathematical models based on the acquired behavior information (step A02).
  • the model learning unit 102 may be configured to execute the learning process every time a preset processing cycle elapses, and to execute the learning process when a predetermined instruction is input by the user. It may be configured.
  • the fitness level calculation unit 103 calculates the fitness level for each of the plurality of mathematical models based on the estimated model parameters (step A03).
  • the fitness level estimation unit 104 determines the fitness level at a future time point for each of the plurality of mathematical models based on time series data including the fitness level calculated at each of a plurality of time points before the current time point. Is estimated (step A04).
  • the model selection unit 105 selects the estimated mathematical model having the highest degree of fitness at a future time from a plurality of mathematical models (step A05).
  • the feature information extraction unit 106 extracts (acquires) feature information by performing random sampling based on the selected mathematical model (step B01).
  • control parameter determination unit 107 estimates the reproduction quality of the content based on the packet based on the extracted feature information. Further, the control parameter determination unit 107 determines a control parameter used when transmitting the packet based on the estimated reproduction quality of the content (step B02). At this time, the control parameter determination unit 107 determines the control parameter so as to maintain or improve the reproduction quality of the content.
  • a mathematical function that can estimate the behavior of an actual packet with high accuracy from a plurality of mathematical models having different numbers of model parameters.
  • a model can be selected. That is, it is possible to specify a mathematical model that can estimate the actual packet behavior with high accuracy. As a result, for example, the behavior of the packet can be estimated with high accuracy using the specified mathematical model.
  • the communication device 100 estimates the fitness at a future time point, and selects a mathematical model based on the estimated fitness. According to this, it is possible to select a mathematical model capable of estimating packet behavior with high accuracy at a future time point. Therefore, by setting the control parameter based on the mathematical model, it is possible to avoid a decrease in the reproduction quality of the content based on the packet. That is, for example, in a service that distributes content data representing content such as video and audio, it is possible to avoid degradation of service quality.
  • the communication system according to the second embodiment includes a plurality of communication terminals including the communication device according to the first embodiment, and is different in that communication is performed between the communication terminals. Accordingly, the following description will focus on such differences.
  • the communication system 1 includes a plurality (two in this example) of communication terminals 110 and 120.
  • the communication terminal 110 and the communication terminal 120 are communicably connected via a communication line NW constituting a communication network.
  • the communication terminal 110 includes a communication device 111 having the same function as the communication device 100 according to the first embodiment, and an application program execution unit 112.
  • the application program execution unit 112 generates content data by executing the application program.
  • the communication terminal 120 includes a communication device 121 having the same function as the communication device 100 according to the first embodiment, and an application program execution unit 122 similar to the application program execution unit 112.
  • the communication terminal 110 is transmitting content data (audio data) containing audio as content to the communication terminal 120. Therefore, the communication terminal 110 is also referred to as a transmission side terminal, and the communication terminal 120 is also referred to as a reception side terminal.
  • the communication terminal 110 transmits a packet (voice packet) including data obtained by dividing the voice data to the communication terminal 120 every time a preset transmission cycle (20 milliseconds in this example) elapses.
  • the communication terminal 120 reproduces the content based on the audio packet received from the communication terminal 110 (that is, outputs audio).
  • the content is audio, but it may be video or video and audio.
  • the communication system 1 operates in the same manner as when content data is transmitted from the communication terminal 110 to the communication terminal 120 when content data is transmitted from the communication terminal 120 to the communication terminal 110.
  • the communication device 121 included in the communication terminal 120 receives a packet transmitted on the communication network by being transmitted by the communication terminal 110. Then, the communication device 121 acquires behavior information based on the received packet. Furthermore, the communication apparatus 121 estimates a model parameter for specifying the mathematical model by performing learning processing on each of the plurality of mathematical models based on the acquired behavior information.
  • the communication device 121 calculates a fitness for each of the plurality of mathematical models based on the estimated model parameters. Further, the communication device 121 estimates the fitness at a future time point for each of the plurality of mathematical models based on time-series data including the fitness values calculated at a plurality of time points before the current time point. To do. Then, the communication device 121 selects the estimated mathematical model having the highest degree of fitness at a future time from a plurality of mathematical models.
  • the communication device 121 extracts (acquires) feature information by performing random sampling based on the selected mathematical model. Then, the communication device 121 estimates the reproduction quality of the content based on the packet based on the extracted feature information. Furthermore, the communication apparatus 121 determines a control parameter used when transmitting a packet based on the estimated reproduction quality of the content.
  • the communication device 121 transmits control parameter information representing the determined control parameter to the communication terminal 110.
  • the communication terminal 110 receives control parameter information from the communication terminal 120, and executes distribution control for transmitting content data to the communication terminal 120 based on the control parameter represented by the received control parameter information.
  • FIG. 5 is an explanatory diagram conceptually showing a hidden Markov model having three states (internal states) S1, S2, and S3.
  • the Hidden Markov Model is a state (internal state) of a target system (in this example, a communication network that transmits a packet) at any point in time, and is based on the state transition probability. It is a model that transitions to a state.
  • the target system when the internal state transitions, the target system outputs one symbol value based on the symbol output probability of the internal state of the transition destination.
  • the symbol value is a value representing the behavior of the packet. Details of the symbol value will be described later.
  • the state of the target system is the internal state S1 at a certain time
  • the state of the target system is one of the three internal states S1, S2, and S3. Transitions with probabilities of 0.3, 0.6, and 0.1, respectively.
  • the target system is based on the symbol output probability of the internal state S2 after the state transitions to the internal state S2.
  • One symbol value is output.
  • the target system outputs any one of the three symbol values a, b, and c with probabilities of 0.3, 0.5, and 0.2, respectively.
  • the symbol value is a discrete value. Therefore, the symbol output probability has a discrete distribution.
  • the symbol value may be a continuous value. In this case, the symbol output probability has a continuous distribution.
  • the hidden Markov model makes it possible to model the behavior of the target system by repeating the transition of the internal state and the output of the symbol value.
  • the number of model parameters is a value corresponding to the number of internal states (the number of internal states).
  • the communication device 121 prepares a plurality of hidden Markov models (mathematical models) with different numbers of internal states (that is, the number of model parameters). Then, the communication device 121 selects an optimal hidden Markov model based on the degree of fitness for each hidden Markov model. As a result, structural changes in the network model (mathematical model) can be captured.
  • the communication system 1 uses information indicating the delay time and the presence / absence of packet loss as the behavior information indicating the behavior of the packet.
  • the dimension of delay time is time
  • the presence or absence of occurrence of packet loss is dimensionless. Therefore, the communication system 1 handles the delay time and the occurrence of packet loss in the same dimension by aligning both dimensions.
  • the communication device 121 converts the acquired behavior information into symbol values corresponding to the magnitude of the delay time and whether or not packet loss has occurred (behavior information conversion means). According to this, the magnitude of the delay time and the presence or absence of occurrence of packet loss can be estimated using the same mathematical model.
  • the communication apparatus 121 divides the magnitude of the delay time for each preset section, and sets different symbol values for each of the plurality of divided sections. assign.
  • the communication device 121 assigns to each section a symbol value that increases as the delay time increases. Furthermore, the communication device 121 assigns a symbol value different from the symbol value assigned to the magnitude of the delay time depending on whether or not packet loss occurs.
  • the communication device 121 converts the acquired behavior information into a symbol value assigned to the behavior of the packet represented by the behavior information (that is, each of the magnitude of the delay time and the occurrence of packet loss). Convert.
  • Non-Patent Document 2 discloses a method of handling in the same dimension as the presence or absence of occurrence of packet loss while keeping the magnitude of the delay time as a continuous value.
  • the communication device 121 may employ this type of method.
  • the communication apparatus 121 estimates the model parameter with respect to the symbol value after conversion.
  • the communication apparatus 121 estimates a model parameter using an algorithm for learning processing such as a Baum-Welch algorithm.
  • the communication device 121 performs the behavior of the packet estimated by the hidden Markov model specified by the estimated model parameter and the behavior of the packet represented by the acquired behavior information. And the degree of matching representing the degree of coincidence.
  • the communication device 121 transmits the symbol value estimated by the hidden Markov model specified by the estimated model parameter and the packet represented by the acquired behavior information.
  • a goodness degree representing the degree of coincidence between the symbol value corresponding to the behavior and the behavior is calculated.
  • the communication device 121 uses the amount of information (entropy) as the fitness.
  • the information amount is a probability that the behavior of the packet represented by the acquired behavior information is generated, and has a value that decreases as the probability based on the hidden Markov model specified by the estimated model parameter increases. That is, the smaller the information amount, the greater the fitness.
  • the information amount S is defined as shown in Equation 2.
  • xt is a symbol value corresponding to the behavior of the packet represented by the acquired behavior information.
  • N is the number of output symbol values, and ⁇ is a model parameter.
  • ⁇ ) is a probability that a symbol value x t will occur when the model parameter is ⁇ .
  • K is the number of internal states.
  • the communication device 121 may be configured to use the information amount S defined as shown in Equation 3 as the fitness.
  • the probability based on the mathematical model that the behavior of the packet represented by the acquired behavior information occurs is the same as the behavior of the packet estimated by the mathematical model and the behavior of the packet represented by the acquired behavior information. It expresses the degree of being well. Therefore, it is preferable to use the amount of information as the fitness as in the communication device 121.
  • the communication device 121 calculates the fitness based on the behavior information acquired in the immediately preceding estimation section T every time the preset estimation section T elapses. .
  • the communication device 121 performs a learning process based on behavior information acquired in the immediately preceding learning section L every time a preset learning section L elapses. Estimate model parameters. Note that the estimation section T and the learning section L may have different lengths or the same length.
  • the communication apparatus 121 is configured to execute the learning process every time the learning section L elapses. However, as described later, the communication apparatus 121 executes the learning process at a timing when a structural change in the network is detected. It may be configured as follows.
  • the lengths of the estimation interval T and the learning interval L are preferably determined according to an empirical rule based on the type of the target network, the type of data transmitted and received by the application program, and the like.
  • the communication device 121 uses model parameters estimated based on behavior information acquired at a time point (previous) before the estimation section T in which behavior information that is a basis for calculating the fitness is acquired. Is preferred.
  • the communication apparatus 121 shown in FIG. 9 the case of calculating the goodness of fit based on the obtained behavior information by estimation interval E t-1 and the estimated interval E t, the learning section L t-1
  • the hidden Markov model specified by the model parameter estimated based on the behavior information acquired in this way is used.
  • the communication apparatus 121 estimates the degree of fitness at a future time point using a time series model (for example, an autoregressive model). Thereafter, the communication apparatus 121 selects a mathematical model having the highest degree of fitness at the future time point from the plurality of mathematical models.
  • a time series model for example, an autoregressive model
  • a value S tot obtained by adding together the fitness S (t) calculated at each of a plurality of (here, m + 1) time points is calculated according to Equation 4.
  • adaptability S (t) is the equation 2, the behavior information acquired by estimation interval E t, a fitness calculated based on.
  • the communication device 121 selects a hidden Markov model having the smallest calculated value S tot .
  • the hidden Markov model having the smallest calculated value S tot is the model that best represents the actual packet behavior.
  • the communication device 121 determines a control parameter for enhancing the reproduction quality of the content based on the selected hidden Markov model.
  • the communication apparatus 121 acquires the frequency (appearance frequency) at which each symbol value appears by performing random sampling based on the selected hidden Markov model.
  • the communication device 121 converts (inversely converts) the symbol value acquired by performing random sampling into the magnitude of the delay time and the presence or absence of occurrence of packet loss corresponding to the symbol value. At this time, the communication device 121 performs the inverse transformation assuming that the probability distribution having the delay time as a random variable is a preset probability distribution (for example, a uniform distribution).
  • the communication apparatus 121 acquires a probability distribution using the magnitude of the delay time as a random variable and a probability distribution using the presence / absence of packet loss as a random variable. Furthermore, the communication apparatus 121 acquires a probability distribution having the number of consecutive packet losses (the number of consecutive packet losses) as a random variable.
  • the communication apparatus 121 uses an evaluation function (for example, ITU-T recommendation G.107 for IP telephones or ITU for videophone services) that takes into consideration the reproduction quality of the content to be reproduced and the influence of the network quality index. -An objective quality estimation method such as T recommendation G.1070) and the acquired probability distribution, control parameters are determined.
  • an evaluation function for example, ITU-T recommendation G.107 for IP telephones or ITU for videophone services
  • error correction processing for example, forward error correction (FEC) processing, etc.
  • FEC forward error correction
  • processing for reproducing (recovering) data included in a packet lost due to occurrence of packet loss is known as processing for reproducing (recovering) data included in a packet lost due to occurrence of packet loss.
  • redundancy of the packet for performing error correction processing that is, a value indicating the ratio of the data amount of the error correction code to the data amount in the packet
  • the content encoding rate is there.
  • the communication apparatus 121 determines a control parameter including the redundancy of the packet and the content encoding rate for performing error correction processing based on a probability distribution using the number of consecutive occurrences of packet loss as a random variable. To do.
  • FIG. 11 is a graph showing a probability distribution with the number of consecutive occurrences of packet loss as a random variable for a plurality of (in this case, three) redundancy levels.
  • 11A is a graph for the smallest redundancy
  • FIG. 11B is a graph for the second smallest redundancy
  • FIG. 11C is a graph for the largest redundancy. It is.
  • the communication device 121 calculates a reproduction quality value corresponding to the redundancy based on the number of consecutive occurrences of packet loss for each redundancy, the coding rate, and the evaluation function in the objective quality estimation method.
  • the redundancy with the highest reproduction quality value is determined as a control parameter.
  • the reproduction quality value is a value representing the reproduction quality of the content.
  • the communication device 121 may be configured to determine the control parameter using another method. For example, the communication device 121 sets the control parameter based on the relationship between the content reproduction quality weighted based on the probability distribution with the network quality index as a random variable, the packet loss rate, and the coding rate. It may be configured to determine.
  • the packet loss rate is the probability that a packet loss will occur.
  • FIG. 12 is a graph showing the relationship between content reproduction quality, packet loss rate, and encoding rate (content reproduction rate) based on the objective quality estimation method.
  • the packet loss rate in this graph is uniform (constant).
  • FIG. 12B is a graph showing the relationship between the content reproduction quality weighted based on the probability distribution with the network quality index as a random variable, the packet loss rate, and the coding rate. is there.
  • the contract clause of the content playback quality guarantee (Service Level Agreement) including the content of compensation when the playback quality of the content is guaranteed and the playback quality cannot be provided Assume that exists.
  • the communication device 121 calculates the probability that the reproduction quality of the content falls below the reference value defined in the content reproduction quality guarantee in FIG. Further, the communication device 121 compares the cost required to improve the reproduction quality of the content with the cost required for the loss of opportunity due to the implementation of the compensation content or the improvement of the reproduction quality based on the calculated probability. . Then, the communication device 121 does not execute the reproduction quality control based on the control parameter when the cost required for implementing the compensation contents or the loss of opportunity is lower than the cost required for enhancing the reproduction quality of the content.
  • the cost required to improve the reproduction quality of content represents an increase in communication bandwidth (network bandwidth) by increasing redundancy without changing the encoding rate.
  • the cost associated with the implementation of the compensation content represents the compensation amount when the reproduction quality of the content cannot be maintained.
  • the cost associated with the lost opportunity represents the revenue lost due to the inability to provide other services due to the increase in the network bandwidth for improving the reproduction quality of the content.
  • the same operations and effects as the communication device 100 according to the first embodiment can be achieved.
  • the communication system according to the third embodiment includes a plurality of relay devices including the communication device according to the first embodiment with respect to the communication system according to the second embodiment so that the relay device specifies a mathematical model. It is different in that it is configured. Accordingly, the following description will focus on such differences.
  • the communication system 2 includes a plurality (two in this example) of relay apparatuses 201 and 211.
  • the relay device 201 and the relay device 211 are communicably connected via a communication line NW1 constituting a communication network.
  • NW1 constituting a communication network.
  • Each relay device 201, 211 includes a communication device having the same function as the communication device 100 according to the first embodiment.
  • the communication system 2 includes a video terminal (communication terminal) 202, a voice terminal (communication terminal) 203, a video terminal (communication terminal) 212, and a voice terminal (communication terminal) 213.
  • the relay device 201, the video terminal 202, and the audio terminal 203 are connected to be communicable via a communication line NW2 that constitutes a communication network.
  • the relay device 211, the video terminal 212, and the audio terminal 213 are communicably connected via a communication line NW3 that constitutes a communication network.
  • the video terminal 202 and the audio terminal 203 are connected to the communication line NW2 to which the relay device 201 is connected. Further, the behavior of packets transmitted on the communication network configured by the same communication line NW2 is similar to each other.
  • the relay apparatus 201 acquires behavior information indicating the behavior of packets transmitted and received by the video terminal 202 and the audio terminal 203, specifies a mathematical model based on the acquired behavior information, Control parameters are determined based on the identified mathematical model. Then, the relay device 201 transmits the determined control parameter to each of the video terminal 202 and the audio terminal 203. Further, the relay device 211 operates in the same manner as the relay device 201.
  • the relay apparatus 201 transmits information representing the specified mathematical model or feature information acquired based on the specified mathematical model to each of the video terminal 202 and the audio terminal 203 without determining the control parameter. It may be configured to. In this case, each of the video terminal 202 and the audio terminal 203 determines a control parameter based on the received information.
  • the same operations and effects as the communication apparatus 100 according to the first embodiment can be achieved.
  • each communication terminal 202, 203, 212, 213 does not need to perform a process for specifying a mathematical model. That is, the processing load on each communication terminal 202, 203, 212, 213 can be reduced. Furthermore, according to the communication system 2 according to the third embodiment, it is possible to increase the number of packets that are the basis of the acquired behavior information, so it is possible to specify a mathematical model with high accuracy.
  • the communication system according to the fourth embodiment includes a resource management server including the communication device according to the first embodiment with respect to the communication system according to the second embodiment, and the resource management server specifies a mathematical model. It is different in that it is configured. Accordingly, the following description will focus on such differences.
  • the communication system 3 includes a resource management server 300, a plurality (two in this example) of wireless access points 301 and 302, and a plurality (four in this example).
  • the resource management server 300, the wireless access point 301, and the wireless access point 302 are communicably connected via a communication line NW4 that constitutes a communication network.
  • Each of the communication terminals 303, 304, 305, 306 is arranged at a position where wireless communication can be performed with both the wireless access point 301 and the wireless access point 302.
  • Each communication terminal 303, 304, and 305 is connected to the wireless access point 301 so that wireless communication is possible.
  • the communication terminal 306 is connected to the wireless access point 302 so as to be able to perform wireless communication.
  • the resource management server 300 includes a communication device having the same function as the communication device 100 according to the first embodiment.
  • Each communication terminal 303, 304, 305, 306 acquires behavior information and transmits the acquired behavior information to the resource management server 300.
  • the resource management server 300 receives behavior information for each of the wireless access points 301 and 302, and specifies a mathematical model based on the received behavior information.
  • the resource management server 300 determines, for each of the wireless access points 301 and 302, whether or not the state of the wireless access point is in a congestion state based on the specified mathematical model. Next, the resource management server 300 connects the communication terminal connected to the wireless access point whose state is determined to be in a congested state so that it can wirelessly communicate with the wireless access point whose state is determined not to be in a congested state. .
  • the resource management server 300 causes the communication terminal 305 to 302 is connected to be wirelessly communicable.
  • the communication system 3 in a situation where a plurality of wireless access points (resources) can be used, it is more appropriate to predict the usage situation of each resource. Resources can be used.
  • the communication system 3 can be configured to prepare a plurality of network paths and use a network path having the highest communication quality.
  • each communication terminal 303, 304, 305, 306 includes a communication device having the same function as the communication device 100 according to the first embodiment. May be.
  • each communication terminal 303, 304, 305, 306 specifies a mathematical model based on the behavior information, and information indicating the specified mathematical model, or feature information acquired based on the specified mathematical model, Transmit to the resource management server 300.
  • the communication device according to the fifth embodiment is configured to generate a packet so as to reproduce the behavior of the packet based on the specified mathematical model with respect to the communication device according to the first embodiment. It is different in point. Accordingly, the following description will focus on such differences.
  • FIG. 15 is a block diagram illustrating functions of the communication device 500 according to the fifth embodiment.
  • the function of the communication device 500 includes a network state estimation unit 500A and a packet generator unit 500B.
  • the network state estimation unit 500A estimates model parameters for specifying a mathematical model for each of a plurality of network models (mathematical models) based on behavior information representing the behavior of packets transmitted over the communication network. To do. Furthermore, the network state estimation unit 500A calculates the fitness for each of the plurality of mathematical models, and selects a mathematical model that can estimate the behavior of the packet with the highest accuracy based on the calculated fitness.
  • the packet generator unit 500B generates a packet by performing random sampling based on the mathematical model selected by the network state estimation unit 500A.
  • the network state estimation unit 500A includes a behavior information acquisition unit (behavior information acquisition unit) 501, a model learning unit (model parameter estimation unit) 502, a fitness level calculation unit (a fitness level calculation unit) 503, and a model selection unit (model). Selection means) 505.
  • a behavior information acquisition unit (behavior information acquisition unit) 501
  • a model learning unit model parameter estimation unit
  • a fitness level calculation unit a fitness level calculation unit
  • model selection unit model selection means
  • the behavior information acquisition unit 501 acquires behavior information indicating the behavior of a packet transmitted on the communication network.
  • the model learning unit 502 estimates, for each of the plurality of mathematical models, model parameters for specifying the mathematical model based on the behavior information acquired by the behavior information acquisition unit 501.
  • the plurality of mathematical models are mathematical models having different numbers of model parameters.
  • Each mathematical model is a mathematical model for estimating packet behavior.
  • the model learning unit 102 estimates model parameters by performing a learning process.
  • the mathematical model is a hidden Markov model. Therefore, the number of model parameters changes according to the number of internal states, which is the number of states in the hidden Markov model.
  • the model parameters include a state transition probability, a symbol output probability, and an initial state probability.
  • the mathematical model may be a time series model (such as an autoregressive moving average model, an autoregressive model, or a moving average model).
  • the number of model parameters is the order in the time series model.
  • the fitness calculation unit 503 calculates the fitness for each of the plurality of mathematical models.
  • the model selection unit 505 selects one mathematical model from a plurality of mathematical models based on the fitness calculated by the fitness calculation unit 503. In this example, the model selection unit 505 selects the mathematical model having the highest calculated fitness.
  • the packet generator unit 500B includes a model information storage unit 506 and a packet generation unit (packet generation unit) 507.
  • the model information storage unit 506 includes model selection information representing the mathematical model selected by the model selection unit 505, model parameters estimated for specifying the mathematical model, and date / time representing the date / time when the mathematical model was selected. Information is stored in association with each other.
  • the packet generation unit 507 is based on the mathematical model represented by the model selection information stored in the model information storage unit 506 and the model parameters stored in the model information storage unit 506 in association with the model selection information. Behavior information is acquired by performing random sampling. Then, the packet generation unit 507 generates a packet based on the acquired behavior information. At this time, the packet generation unit 507 generates a packet at a timing corresponding to the date and time represented by the date and time information stored in the model information storage unit 506 in association with the model selection information.
  • the packet generation unit 507 performs random sampling based on the mathematical model selected by the model selection unit 505 and the model parameter estimated by the model learning unit 502 at each of a plurality of time points.
  • the packet is generated so as to reproduce the behavior of the packet at each of the plurality of time points.
  • the operation of the communication apparatus 500 according to the present invention includes an operation of selecting a mathematical model as shown in FIG. 16 and an operation of generating a packet as shown in FIG. First, the operation of selecting a mathematical model will be described with reference to FIG.
  • Communication device 500 receives a packet transmitted over a communication network.
  • the behavior information acquisition unit 501 acquires behavior information based on the received packet (step C01).
  • the model learning unit 502 estimates a model parameter for specifying the mathematical model by performing a learning process on each of the plurality of mathematical models based on the acquired behavior information (step C02).
  • the fitness calculator 503 calculates fitness for each of the plurality of mathematical models based on the estimated model parameters (step C03). Then, the model selection unit 505 selects a mathematical model having the highest calculated fitness from a plurality of mathematical models (step C04).
  • the model information storage unit 506 includes model selection information representing the selected mathematical model, model parameters estimated to identify the mathematical model, date / time information representing the date and time when the mathematical model was selected, Are stored in association with each other (step C05).
  • the packet generation unit 507 is based on the mathematical model represented by the model selection information stored in the model information storage unit 506 and the model parameters stored in the model information storage unit 506 in association with the model selection information. Behavior information is acquired by performing random sampling (step D01).
  • the packet generation unit 507 Based on the acquired behavior information, the packet generation unit 507 generates a packet at a timing corresponding to the date and time indicated by the date and time information stored in the model information storage unit 506 in association with the model selection information. (Step D02).
  • the packet generation unit 507 repeatedly executes the processing of step D01 and step D02 according to the date and time represented by the date and time information stored in the model information storage unit 506.
  • a packet is generated based on a mathematical model capable of estimating the behavior of a packet with high accuracy at each of a plurality of time points. be able to. As a result, the past packet behavior can be reproduced with high accuracy.
  • the communication system according to the sixth embodiment is different in that it includes a plurality of packet generators having the function of the communication device according to the fifth embodiment. Accordingly, the following description will focus on such differences.
  • the communication system 6 includes a plurality (two in this example) of packet generators 601 and 605, a transmission device 602, and an evaluation device 606. Note that the packet generator 601 and the packet generator 605 may be configured by the same device.
  • the packet generator 601 and the transmission device 602 are communicably connected via a communication line NW5 that constitutes a communication network.
  • the packet generator 605 and the evaluation device 606 are communicably connected via a communication line NW6 constituting a communication network.
  • the packet generator 601 and the packet generator 605 are connected so as to be communicable.
  • the transmission device 602 transmits a packet based on content data (for example, audio data) to the packet generator 601.
  • content data for example, audio data
  • the packet generator 601. In this example, in the communication network configured by the communication line NW5, cross traffic is generated or traffic flow control is performed.
  • the packet generator 601 acquires behavior information indicating the behavior of the packet transmitted by the transmission device 602.
  • the packet generator 601 estimates a model parameter for specifying the mathematical model by performing learning processing on each of the plurality of mathematical models based on the acquired behavior information.
  • the packet generator 601 calculates the fitness for each of the plurality of mathematical models based on the estimated model parameters. Then, the packet generator 601 selects a mathematical model having the highest calculated fitness from a plurality of mathematical models.
  • the packet generator 601 associates model selection information representing the selected mathematical model, model parameters estimated for specifying the mathematical model, and date / time information representing the date and time when the mathematical model was selected. Add and remember.
  • the transmission device 602 may be configured to change the control parameter based on the mathematical model selected by the packet generator 601.
  • the packet generator 601 transmits all stored information to the packet generator 605 as a data set.
  • the packet generator 605 generates a packet by performing random sampling based on the received data set.
  • the packet generator 605 transmits the generated packet to the evaluation device 606 via the communication line NW6.
  • the evaluation device 606 receives the packet transmitted by the packet generator 605.
  • the evaluation device 606 evaluates content reproduction quality (for example, audio distribution reproduction quality) based on the received packet.
  • the evaluation device 606 evaluates the change in the reproduction quality of the content due to the change of the control parameter. It is preferred to do so.
  • the same operations and effects as the communication device 500 according to the fifth embodiment can be achieved.
  • the model specifying apparatus 700 is A behavior information acquisition unit (behavior information acquisition means) 701 for acquiring behavior information indicating the behavior of a packet transmitted over a communication network; For each of a plurality of mathematical models specified by model parameters and for estimating the behavior of the packet, the number of the model parameters being different, the acquired behavior information A model parameter estimation unit (model parameter estimation means) 702 for estimating a model parameter for specifying the mathematical model based on For each of the plurality of mathematical models, a degree of fitness representing the degree to which the behavior of the packet estimated by the mathematical model matches the behavior of the packet represented by the acquired behavior information is calculated. A fitness level calculation unit (a fitness level calculation means) 703, A model selection unit (model selection means) 704 that selects one mathematical model from the plurality of mathematical models based on the calculated fitness; Is provided.
  • a fitness level calculation unit a fitness level calculation means
  • a model selection unit (model selection means) 704 that selects one mathematical model from the plurality of mathematical models based on the calculated fitness; Is provided.
  • a mathematical model capable of estimating the actual packet behavior with high accuracy can be selected from a plurality of mathematical models having different numbers of model parameters. That is, it is possible to specify a mathematical model that can estimate the actual packet behavior with high accuracy. As a result, for example, the behavior of the packet can be estimated with high accuracy using the specified mathematical model.
  • each function of each device is realized by the CPU executing a program (software), but may be realized by hardware such as a circuit.
  • the program is stored in the storage device, but may be stored in a computer-readable recording medium.
  • the recording medium is a portable medium such as a flexible disk, an optical disk, a magneto-optical disk, and a semiconductor memory.
  • Behavior information acquisition means for acquiring behavior information representing the behavior of a packet transmitted over a communication network; For each of a plurality of mathematical models that are specified by model parameters and that estimate the behavior of the packet and that have different numbers of model parameters, the acquired behavior information is included in the acquired behavior information.
  • a model parameter estimating means for estimating a model parameter for identifying the mathematical model, For each of the plurality of mathematical models, a degree of fitness representing the degree of coincidence between the behavior of the packet estimated by the mathematical model and the behavior of the packet represented by the acquired behavior information is calculated.
  • a fitness calculation means for Model selection means for selecting one mathematical model from the plurality of mathematical models based on the calculated fitness;
  • a model specifying device comprising:
  • a mathematical model capable of estimating the actual packet behavior with high accuracy can be selected from a plurality of mathematical models having different numbers of model parameters. That is, it is possible to specify a mathematical model that can estimate the actual packet behavior with high accuracy. As a result, for example, the behavior of the packet can be estimated with high accuracy using the specified mathematical model.
  • the model specifying device (Appendix 2) The model specifying device according to attachment 1, wherein The mathematical model is a hidden Markov model,
  • the number of model parameters is a value according to the number of internal states that is the number of states in the hidden Markov model,
  • the model parameters include a state transition probability that is a probability of transition of the state, a symbol output probability that is a probability distribution having a symbol value output when the state transitions as a random variable, and a probability of the state at an initial stage.
  • a model specifying device including an initial state probability which is a probability distribution as a variable.
  • the model specifying device (Appendix 3) The model specifying device according to attachment 2, wherein
  • the behavior information is a delay time that is a time required for the packet to reach the transmission destination from the transmission source, and information indicating whether or not a packet loss that does not reach the transmission destination of the packet occurs.
  • Behavior information conversion means for converting the acquired behavior information into the symbol values corresponding to the magnitude of the delay time and the presence or absence of the packet loss,
  • the model specifying device configured to estimate the model parameter for the symbol value after the conversion.
  • the model specifying device according to any one of Appendix 1 to Appendix 3,
  • the fitness level calculating unit is configured to use, as the fitness level, an information amount having a value that decreases as the probability based on the mathematical model, in which the behavior of the packet represented by the acquired behavior information occurs, increases.
  • Model specific device is configured to use, as the fitness level, an information amount having a value that decreases as the probability based on the mathematical model, in which the behavior of the packet represented by the acquired behavior information occurs, increases.
  • the probability based on the mathematical model that the behavior of the packet represented by the acquired behavior information occurs is the same as the behavior of the packet estimated by the mathematical model and the behavior of the packet represented by the acquired behavior information. It expresses the degree of being well. Therefore, it is preferable to use the amount of information as the fitness, as in the model specifying device.
  • the model specifying device according to any one of Appendix 1 to Appendix 4, For each of the plurality of mathematical models, a fitness level estimation means for estimating a fitness level at a future time point based on time series data composed of fitness levels calculated at each of a plurality of time points, The model specifying device is configured to select one mathematical model from the plurality of mathematical models based on the estimated future fitness.
  • control parameter is, for example, the redundancy that the packet has for performing error correction processing, and / or the encoding rate that is used when encoding the content represented by the packet.
  • a model specifying apparatus comprising: a statistic calculating unit that calculates a statistic of the behavior of the packet by performing random sampling based on the selected mathematical model.
  • Appendix 7 The model specifying device according to any one of appendices 1 to 6, A model specifying device comprising reproduction quality estimation means for estimating reproduction quality of content based on the packet based on the selected mathematical model.
  • Appendix 8 The model specifying device according to any one of Appendix 1 to Appendix 7, A model specifying device comprising control parameter determining means for determining a control parameter used when transmitting the packet based on the selected mathematical model.
  • a model specifying device (Appendix 10) The model specifying device according to any one of Appendix 1 to Appendix 4, In order to reproduce the behavior of the packet at each of the plurality of time points by performing random sampling based on the selected mathematical model and the estimated model parameter at each of the plurality of time points.
  • a model specifying apparatus comprising packet generation means for generating the packet.
  • a packet can be generated based on a mathematical model capable of estimating the behavior of the packet with high accuracy. As a result, the past packet behavior can be reproduced with high accuracy.
  • (Appendix 11) Obtain behavior information that represents the behavior of packets transmitted over the communication network, For each of a plurality of mathematical models that are specified by model parameters and that estimate the behavior of the packet and that have different numbers of model parameters, the acquired behavior information is included in the acquired behavior information. Based on the model parameters for identifying the mathematical model, For each of the plurality of mathematical models, a degree of fitness representing the degree of coincidence between the behavior of the packet estimated by the mathematical model and the behavior of the packet represented by the acquired behavior information is calculated. And A model specifying method of selecting one mathematical model from the plurality of mathematical models based on the calculated fitness.
  • the model specifying method according to appendix 11, wherein The mathematical model is a hidden Markov model,
  • the number of model parameters is a value according to the number of internal states that is the number of states in the hidden Markov model,
  • the model parameters include a state transition probability that is a probability of transition of the state, a symbol output probability that is a probability distribution having a symbol value output when the state transitions as a random variable, and a probability of the state at an initial stage.
  • a model specifying method including an initial state probability which is a probability distribution as a variable.
  • the behavior information is a delay time that is a time required for the packet to reach the transmission destination from the transmission source, and information indicating whether or not a packet loss that does not reach the transmission destination of the packet occurs.
  • the acquired behavior information is converted into the symbol value corresponding to each of the magnitude of the delay time and the presence or absence of the packet loss,
  • a model identification method configured to estimate the model parameter for the converted symbol value.
  • Behavior information acquisition means for acquiring behavior information representing the behavior of a packet transmitted over a communication network; For each of a plurality of mathematical models that are specified by model parameters and that estimate the behavior of the packet and that have different numbers of model parameters, the acquired behavior information is included in the acquired behavior information.
  • a model parameter estimating means for estimating a model parameter for identifying the mathematical model, For each of the plurality of mathematical models, a degree of fitness representing the degree of coincidence between the behavior of the packet estimated by the mathematical model and the behavior of the packet represented by the acquired behavior information is calculated.
  • a fitness calculation means for Model selection means for selecting one mathematical model from the plurality of mathematical models based on the calculated fitness;
  • a model identification program for realizing
  • the mathematical model is a hidden Markov model
  • the number of model parameters is a value according to the number of internal states that is the number of states in the hidden Markov model
  • the model parameters include a state transition probability that is a probability of transition of the state, a symbol output probability that is a probability distribution having a symbol value output when the state transitions as a random variable, and a probability of the state at an initial stage.
  • a model specifying program including an initial state probability which is a probability distribution as a variable.
  • the model specifying program according to attachment 15, wherein
  • the behavior information is a delay time that is a time required for the packet to reach the transmission destination from the transmission source, and information indicating whether or not a packet loss that does not reach the transmission destination of the packet occurs.
  • Realizing behavior information conversion means for converting the acquired behavior information into the symbol values corresponding to the magnitude of the delay time and the presence or absence of the packet loss
  • the model parameter estimating unit is configured to estimate the model parameter with respect to the converted symbol value.
  • the present invention is applicable to a model specifying device that specifies a mathematical model for estimating packet behavior, a content distribution system that transmits and receives content data, and the like.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

La présente invention se rapporte à un dispositif de spécification de modèle (700) qui est pourvu : d'une unité d'acquisition d'informations de comportement (701) pour acquérir des informations de comportement qui représentent un comportement de paquet ; d'une unité d'estimation des paramètres de modèle (702) pour estimer des paramètres de modèle qui spécifient un modèle mathématique, ladite estimation étant effectuée sur la base des informations de comportement acquises et pour chaque modèle mathématique d'une pluralité de modèles mathématiques qui sont utilisés pour estimer le comportement de paquet et ont un nombre différent de paramètres de modèle ; d'une unité de calcul du degré de conformité (703) pour calculer pour chaque modèle mathématique de la pluralité de modèles mathématiques un degré de conformité qui représente le degré de concordance entre le comportement de paquet estimé par le modèle mathématique et le comportement de paquet représenté par les informations de comportement acquises ; et d'une unité de sélection de modèle (704) pour sélectionner un modèle mathématique parmi la pluralité de modèles mathématiques sur la base du degré de conformité calculé.
PCT/JP2011/002304 2010-06-16 2011-04-20 Dispositif de spécification de modèle WO2011158421A1 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP2012520251A JPWO2011158421A1 (ja) 2010-06-16 2011-04-20 モデル特定装置

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2010-137049 2010-06-16
JP2010137049 2010-06-16

Publications (1)

Publication Number Publication Date
WO2011158421A1 true WO2011158421A1 (fr) 2011-12-22

Family

ID=45347839

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2011/002304 WO2011158421A1 (fr) 2010-06-16 2011-04-20 Dispositif de spécification de modèle

Country Status (2)

Country Link
JP (1) JPWO2011158421A1 (fr)
WO (1) WO2011158421A1 (fr)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11438246B2 (en) 2018-03-29 2022-09-06 Nec Corporation Communication traffic analyzing apparatus, communication traffic analyzing method, program, and recording medium
US11509539B2 (en) 2017-10-26 2022-11-22 Nec Corporation Traffic analysis apparatus, system, method, and program

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH1168849A (ja) * 1997-08-12 1999-03-09 Kokusai Denshin Denwa Co Ltd <Kdd> トラヒックジェネレータおよびトラヒック生成関数決定方法
JP2005156593A (ja) * 2003-11-20 2005-06-16 Seiko Epson Corp 音響モデル作成方法、音響モデル作成装置、音響モデル作成プログラムおよび音声認識装置
JP2009188885A (ja) * 2008-02-08 2009-08-20 Nec Corp 通信装置、通信システム、通信方法及び通信プログラム
JP2009194507A (ja) * 2008-02-13 2009-08-27 Nec Corp パケット通信装置及びパケット通信方法並びにパケット通信プログラム

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH1168849A (ja) * 1997-08-12 1999-03-09 Kokusai Denshin Denwa Co Ltd <Kdd> トラヒックジェネレータおよびトラヒック生成関数決定方法
JP2005156593A (ja) * 2003-11-20 2005-06-16 Seiko Epson Corp 音響モデル作成方法、音響モデル作成装置、音響モデル作成プログラムおよび音声認識装置
JP2009188885A (ja) * 2008-02-08 2009-08-20 Nec Corp 通信装置、通信システム、通信方法及び通信プログラム
JP2009194507A (ja) * 2008-02-13 2009-08-27 Nec Corp パケット通信装置及びパケット通信方法並びにパケット通信プログラム

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
KOSUKE NOGAMI ET AL.: "IN2008-93 Switching Control Scheme for Video Streaming based on Predicting Network", IEICE TECHNICAL REPORT, vol. 108, no. 342, 4 December 2008 (2008-12-04), pages 43 - 48 *
KOSUKE NOGAMI ET AL.: "IN2010-53 Network Traffic Estimation with Dynamic Model Selection Approach", IEICE TECHNICAL REPORT, vol. 110, no. 191, 26 August 2010 (2010-08-26), pages 55 - 60 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11509539B2 (en) 2017-10-26 2022-11-22 Nec Corporation Traffic analysis apparatus, system, method, and program
US11438246B2 (en) 2018-03-29 2022-09-06 Nec Corporation Communication traffic analyzing apparatus, communication traffic analyzing method, program, and recording medium

Also Published As

Publication number Publication date
JPWO2011158421A1 (ja) 2013-08-19

Similar Documents

Publication Publication Date Title
Karagkioules et al. Online learning for low-latency adaptive streaming
Zhani et al. Analysis and Prediction of Real Network Traffic.
Yoo Efficient traffic prediction scheme for real-time VBR MPEG video transmission over high-speed networks
González et al. Simulation and experimental testbed for adaptive video streaming in ad hoc networks
Gomes et al. Using fuzzy link cost and dynamic choice of link quality metrics to achieve QoS and QoE in wireless mesh networks
CN111617466B (zh) 编码格式的确定方法、装置及云游戏的实现方法
Keshtkarjahromi et al. Content-aware instantly decodable network coding over wireless networks
WO2011158421A1 (fr) Dispositif de spécification de modèle
Keshtkarjahromi et al. Content-aware network coding over device-to-device networks
US20130243082A1 (en) Rate optimisation for scalable video transmission
Khayam et al. Linear-complexity models for wireless MAC-to-MAC channels
Goudarzi Scalable video transmission over multi-hop wireless networks with enhanced quality of experience using swarm intelligence
WO2017161122A1 (fr) Système de transmission vidéo en continu en direct au moyen de codes fontaines compatibles avec le retard
Gani et al. Prediction of State of Wireless Network Using Markov and Hidden Markov Model.
Radhika et al. Video Traffic Analysis over LEACH-GA routing protocol in a WSN
Rani et al. Hybrid evolutionary computing algorithms and statistical methods based optimal fragmentation in smart cloud networks
Boushaba et al. Intelligent multipath optimized link state routing protocol for QoS and QoE enhancement of video transmission in MANETs
Hwang et al. A novel user-oriented quality of service algorithm for home networks
JP2013030852A (ja) スケーラブル映像符号化装置及び方法及びプログラム
JP5675164B2 (ja) 送信装置、送信方法、並びにプログラム
JP5034998B2 (ja) 通信装置、通信システム、通信方法及び通信プログラム
Al-Sbou et al. A novel quality of service assessment of multimedia traffic over wireless ad hoc networks
Das et al. Coping with uncertainty in mobile wireless networks
Singh et al. Modeling and Optimization of Video Transmission in Data Compression & Source Coding
Nihei et al. A QoE indicator and a transmission control method for VoIP on mobile networks considering delay spikes

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 11795329

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 2012520251

Country of ref document: JP

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 11795329

Country of ref document: EP

Kind code of ref document: A1