JP7410679B2 - Receiving system, communication compensation requesting device, and program - Google Patents

Description

The present invention relates to a receiving system, a communication compensation requesting device, and a program, and more particularly to a receiving system, a communication compensation requesting device, and a program for compensating data through communication when broadcast waves are attenuated.

When digital broadcast waves for satellite broadcasting and terrestrial broadcasting deteriorate due to attenuation due to rain in the case of satellite digital broadcasting and fading in digital terrestrial broadcasting, it becomes difficult to receive the broadcast waves. Regarding rainfall attenuation, the amount of attenuation increases as the frequency increases, so there is concern that reception conditions will further deteriorate in satellite broadcasting using radio waves in the 21 GHz band, which will become available in the future. There is a need for compensation technology for broadcast waves so that content can continue to be viewed even when reception conditions deteriorate due to rain attenuation or fading. When reception of broadcast waves becomes difficult, multiple compensation techniques have been proposed that allow content distribution (compensation data) of the same content to be received via communication using the Internet, etc., so that the content can be viewed continuously ( Patent Documents 1 to 4).

Here, "broadcasting" refers to a transmission method for broadcasting content to an unspecified number of users, and particularly assumes distribution by radio waves. Furthermore, "communication" refers to a transmission method for bidirectionally transmitting and receiving information, and in particular, assumes a means of acquiring data via an IP (Internet Protocol) network or the like to supplement content.

When considering the compensation of broadcast waves through these communications, rather than requesting content via communication after the broadcast waves are completely unreceivable, it is possible to predict the deterioration of the broadcast wave reception situation and request content in advance. It is desirable to be able to continue stable viewing by requesting.

In the reception of satellite broadcasting, there is a document that proposes a mechanism for predicting deterioration in reception conditions, requesting content, and compensating for broadcast waves (Patent Document 5), but does not mention a specific prediction means. In addition, when receiving terrestrial broadcasting on a mobile object, when compensation is provided by communication when moving to a place where reception is difficult, the future latitude and longitude of the mobile object can be predicted from continuous information on the latitude and longitude of the mobile object, and content can be received in advance. A mechanism for requesting this has been proposed (Patent Document 6). However, although the method of predicting the destination is effective for compensating broadcast waves in a mobile device, it is also applicable to predicting deterioration in reception conditions that can occur in fixed reception, such as rain attenuation in satellite broadcasting and fading in terrestrial broadcasting. I can't. The mechanisms for requesting content in advance that have been proposed so far differ from prediction of reception conditions; for example, they constantly measure the reception C/N (carrier-noise ratio) of broadcast waves, and By setting a certain margin for the value of N at which reception becomes impossible, content is requested a certain amount before reception becomes impossible (Patent Document 7).

As a method for predicting time-series changes in the reception status of broadcast waves, prediction of rainfall attenuation for satellite broadcasting has been proposed (Patent Document 8). For example, based on records of 10-minute measurements from a rain gauge placed at an observation point, or 10-minute rainfall data at an observation point from the Japan Meteorological Agency, the future 1-minute rainfall value at that observation point is probabilistically predicted. Based on this prediction, it is mentioned that in areas where reception conditions deteriorate due to rainfall attenuation, the power can be increased using, for example, a phased array antenna to improve reception conditions. However, while the Japan Meteorological Agency's observation points may be sufficient to determine the power boost points for phased array antennas, the number of observation points is insufficient to predict deterioration in reception conditions in each home where each receiving device is installed. Of course, it is difficult to install a rain gauge in each household. Furthermore, when considering communication compensation, it is necessary to predict reception conditions at time intervals shorter than one minute. Furthermore, it cannot cope with fading in digital terrestrial broadcasting.

On the other hand, there are many proposals regarding weather forecasting and forecasting, such as measuring the reception strength of broadcasting satellites and communication satellites, measuring weather information such as the amount of clouds from the measured reception strength, and making weather predictions based on the measurement results. A method for implementing this has been proposed (Patent Document 9). However, the purpose of these methods is weather prediction, not prediction of the reception status of broadcast waves.

A method using a multilayer perceptron (Non-patent Document 1) or ARIMA, a statistical time series model used in finance, is used to predict rainfall attenuation of radio waves in the Ka band, including the 21 GHz band, which is being considered for future satellite broadcasting. A method using the /GARCH model (Non-Patent Document 2) has been proposed. These proposals are aimed at predicting the rainfall attenuation of radio waves using a functional method and using it for the aforementioned phased array antenna and site diversity at transmitting stations.

Japanese Patent Application Publication No. 2004-40380 Japanese Patent Application Publication No. 2004-220569 Japanese Patent Application Publication No. 2017-199956 International Publication No. 2015/048569 Japanese Patent Application Publication No. 2012-138649 Japanese Patent Application Publication No. 2007-259049 Japanese Patent Application Publication No. 2003-134064 Japanese Patent Application Publication No. 2006-246375 Japanese Patent Application Publication No. 2005-318398

A. P. Chambers and I. E. Otung, “Neural network approach to short-term fade prediction on satellite links,” Electron. Lett., (2005), vol. 41, no. 23, pp. 45-46 L. De Montera, C. Mallet, L. Barthes, and P. Gole, “Short-term prediction of rain attenuation level and volatility in Earth-to-Satellite links at EHF band,” Nonlinear Process. Geophys., (2008) , vol. 15, no. 4, pp. 631-643 F. Chollet et al., “Keras,” Internet <URL: https://keras.io>

In order to stably compensate for the deterioration of digital broadcast reception conditions through communication, it is necessary to predict the reception conditions of digital broadcasts. However, as mentioned above, previous proposals do not have a mechanism to predict time-series deterioration of reception conditions such as rain attenuation and fading, and in some cases, content may be lost, or even if there is no loss, it may be difficult to connect to the communication. A situation arises where you have to wait for playback.

By using a threshold with an appropriate margin for the value that makes reception impossible, an effect similar to predicting the deterioration of reception conditions can be obtained, but for example, in the case of rainfall attenuation, In some cases, the amount of rain suddenly increases and it is not possible to request content before it decays, or if a pessimistic threshold is prepared, content is constantly requested from the communication, etc. It is difficult to prepare a threshold.

Furthermore, the method of predicting time-series changes in rainfall attenuation using rainfall data, etc. is premised on application to transmitting devices, and is implemented in each receiving device, for example, a TV in each home, to predict changes in reception conditions. That's difficult. Although methods have been proposed for predicting weather forecasting and deterioration of reception conditions, no method has been proposed that utilizes time-series data of received radio waves to predict time-series changes in reception conditions and apply it to receiving equipment. .

Therefore, an object of the present invention, which was made in view of the above-mentioned problems, is to predict time-series changes in reception conditions by inputting only the state of received radio waves that can be measured by each receiving device (reception system), and to estimate the reception conditions. It is an object of the present invention to provide a receiving system, a communication compensation requesting device, and a program that allow stable content viewing to continue even when the situation deteriorates.

In order to solve the above problems, a communication compensation requesting device according to the present invention is a communication compensation requesting device used in a receiving system that receives compensation data by communication when a broadcast wave is attenuated. A recording unit that records the quality and outputs a received radio wave quality data string, and a recurrent neural network (RNN), and uses the received radio wave quality data string as input to predict the received radio wave quality after a predetermined time, or a prediction unit that predicts a determination as to whether the compensation data should be requested after a predetermined time and outputs a request signal for the compensation data based on the prediction; and a learning data generation unit that generates learning data from the received radio wave quality data string. and the recurrent neural network performs learning to predict reception radio wave quality after a predetermined time or a determination as to whether compensation data should be requested after a predetermined time based on the learning data in a learning mode, The learning data generation unit includes a storage unit that stores the received radio wave quality data string, and uses the received radio wave quality data string to generate a range that is effective as learning data, including received radio wave quality that makes it impossible to receive broadcast waves. a determining unit for determining, and a generating unit for generating learning data by selecting the received radio wave quality data string in the range determined by the determining unit from the received radio wave quality data string in the storage unit, The learning data includes a received radio wave quality data string that is input data, a correct received radio wave quality that is the received radio wave quality after a predetermined time from the time when the received radio wave quality data string is input, and the received radio wave quality data string. At least one of the correct compensation request determinations that determine whether compensation data should be requested after a predetermined time from the point in time is set as a set, and the determining unit checks the latest received radio wave quality data among the received radio wave quality data string, There is a time when the latest received radio wave quality data becomes unreceivable, and there is a first point in time when the normal received radio wave quality becomes degraded received radio wave quality, and a time when the degraded received radio wave quality becomes normal. A second time point when the received radio wave quality returns to normal is specified, and a predetermined margin is added to the range between the first time point and the second time point to make the range effective as the learning data. shall be.

The communication compensation requesting device may further include a buffer for storing a predetermined number of received radio wave quality data, wherein the recording unit records the received radio wave quality at regular intervals, and the recording unit records the received radio wave quality data at regular intervals. It is desirable to output the received radio wave quality data as a received radio wave quality data string.

Further, in the communication compensation requesting device, the prediction unit includes the recurrent neural network therein, and predicts the received radio wave quality after a predetermined time by inputting the received radio wave quality data string, or predicts the received radio wave quality after a predetermined time. an RNN unit that predicts and outputs a determination as to whether compensation data should be requested; the latest data of the input received radio wave quality data string; and a prediction of the received radio wave quality output from the RNN unit or the compensation data. It is preferable to include a compensation data requesting unit that outputs a request signal for the compensation data to the compensation data server based on a prediction of whether to request the compensation data.

In order to solve the above problem, a program according to the present invention is characterized in that it causes a computer to function as the communication compensation requesting device.
Further, in order to solve the above problems, a receiving system according to the present invention includes a broadcast wave receiving device that receives, demodulates and decodes broadcast waves, and the communication compensation requesting device, which A communication compensation requesting device that uses a recurrent neural network (RNN) to predict the reception status of broadcast waves after a predetermined time of reception based on series data, and requests compensation data by communication when the reception status is predicted to deteriorate. a compensation data receiving device that receives and decodes the compensation data obtained in response to a request from the communication compensation requesting device; and a compensation data receiving device that receives and decodes the compensation data obtained in response to a request from the communication compensation requesting device; The present invention is characterized by comprising a broadcast wave compensation data switch that switches and outputs the broadcast wave reception data and the compensation data output from the compensation data receiving device.

According to the receiving system, communication compensation requesting device, and program of the present invention, only the state of received radio waves that can be measured by each receiving device (receiving system) is used as input to predict time-series changes in the receiving situation, and the receiving situation deteriorates. Even when this happens, you can continue to enjoy stable content viewing.

1 is a diagram showing an example of an overall diagram of a reception system of the present invention. 1 is a diagram showing an example of the configuration of a communication compensation requesting device according to the present invention. FIG. 3 is a diagram illustrating an example of creating a received radio wave quality data string. FIG. 3 is a diagram showing an example of the configuration of a learning data generation section. 3 is an example of a flowchart showing an operation algorithm of a determination unit. FIG. 3 is a diagram illustrating how to determine the effective range of learning data. FIG. 3 is a diagram illustrating an example of creating learning data. It is a figure showing an example of composition of a prediction part. 7 is an example of a flowchart showing an operation algorithm of a compensation data requesting section. 7 is another example of a flowchart showing the operation algorithm of the compensation data requesting section. FIG. 6 is a diagram showing another example of the configuration of the communication compensation requesting device of the present invention. It is a diagram showing the beacon radio wave reception status in the 21 GHz band on August 3, 2018. FIG. 6 is a diagram showing a range of reception status data determined to be valid as learning data. It is a diagram showing the beacon radio wave reception situation in the 21 GHz band on September 4, 2018. FIG. 3 is a diagram illustrating changes in predicted received radio wave quality and compensation data request period. FIG. 7 is a diagram showing a compensation data request period when a margin is provided for the threshold value. FIG. 7 is a diagram illustrating a compensation data request period based on a prediction of reception status when a threshold value is adjusted.

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 shows an example of an overall diagram of a receiving system of the present invention. The receiving system 100 includes a broadcast wave receiving device 10, a communication compensation requesting device 20, a compensation data receiving device 30, and a broadcast wave compensation data switch 40. The receiving system 100 may further include a playback device 50.

A satellite broadcast received by a satellite antenna 61 and/or a terrestrial broadcast received by a terrestrial antenna 62 are input to the broadcast wave receiving device 10 as broadcast waves. The broadcast wave receiving device 10 demodulates and decodes the received broadcast wave, and outputs the decoded program (content) to the broadcast wave compensation data switch 40 as broadcast wave reception data. In a normal reception state, the program decoded by the broadcast wave receiving device 10 is input to the playback device 50 via the switch 40 and played back. When the reception condition deteriorates to the extent that the broadcast wave cannot be reproduced, the broadcast wave receiving device 10 outputs a switching signal to the broadcast wave compensation data switch 40.

The received broadcast waves are also input to the communication compensation requesting device 20, and the reception status is analyzed. Here, the reception condition is the condition of the received radio waves of the broadcast wave, and is measured using the received electric field value, MER (Modulation Error Ratio), BER (Bit Error Rate), C/N, etc. This is data that can be detected quantitatively. The communication compensation requesting device 20 uses a recurrent neural network (RNN) to detect time-series changes in the reception status (reception status of the broadcast wave after a predetermined time of reception) based on time-series data of the reception status of the broadcast waves. ) to predict. When a deterioration of the reception condition is predicted, a request signal requesting compensation data is outputted to the compensation data server 120 through communication. As will be described later, the learning of the recurrent neural network can be performed inside the communication compensation requesting device 20, but it is also possible to transmit the recording data of the reception situation to the external learning server 110 and perform the learning externally. However, it is also possible to receive a trained recurrent neural network model (RNN model) constructed by the learning server 110 and reflect it in the communication compensation requesting device 20.

The compensation data receiving device 30 receives the compensation data received via communication from the compensation data server 120 as a response to the request signal from the communication compensation requesting device 20, decodes it, and outputs it to the broadcast wave compensation data switch 40. .

The broadcast wave compensation data switch 40 receives the program (broadcast wave reception data) decoded by the broadcast wave reception device 10 and the compensation data decoded by the compensation data reception device 30, and sends one of them to the playback device 50. Output. In normal reception conditions, the broadcast wave compensation data switch 40 outputs the broadcast wave data to the reproducing device 50, but if the reception condition deteriorates to the extent that the broadcast waves cannot be reproduced, a switching signal from the broadcast wave receiving device 10 is output. The switching is performed based on the compensation data and the compensation data is output to the playback device 50. Note that after the reception status is restored, the broadcast wave data is outputted to the reproducing device 50 again based on the switching signal.

The playback device 50 is a general display device, audio output device, etc., and plays back the data input from the broadcast wave compensation data switch 40. Based on the operation of the switch 40, broadcast wave reception data is normally reproduced, but when reception conditions are poor, compensation data is reproduced.

Next, the communication compensation requesting device 20 will be explained in detail. FIG. 2 shows an example of the configuration of the communication compensation requesting device 20. In this embodiment, the communication compensation requesting device 20 includes a recording section 21, a learning data generation section 22, and a prediction section 23.

The recording unit 21 receives broadcast waves from the receiving antennas 61 and 62, analyzes the broadcast waves, and records received radio wave quality such as received electric field value, MER, and C/N. The received electric field value, MER, C/N, etc. may be recorded individually, or any one of them may be recorded and used for predicting the reception situation. Note that the received electric field value can be used as data for determining whether or not it is possible to reproduce a program on a broadcast wave, so in this specification, the received electric field value is also treated as one type of received radio wave quality. As described later, the recording unit 21 creates a received radio wave quality data string and outputs it to the learning data generation unit 22 and the prediction unit 23 at regular intervals, that is, every unit time (sampling time) t.

The learning data generating section 22 receives the received radio wave quality data string output from the recording section 21 and generates learning data for the recurrent neural network to perform learning. The generated learning data is output to the prediction unit 23.

The prediction unit 23 includes a recurrent neural network (RNN) therein, predicts time-series changes in the reception status of broadcast waves, and outputs a compensation data request signal when a deterioration of the reception status is predicted. The operation of the prediction unit 23 is divided into operation in learning mode and operation in prediction mode. In the learning mode, the recurrent neural network is trained based on the learning data output from the learning data generation section 22, and parameters are optimized. In the prediction mode, the received radio wave quality data string inputted from the recording unit 21 is used to predict the reception status (received radio wave quality) after a predetermined time or the determination whether to request compensation data after a predetermined time. If it is determined from the prediction result that a request for compensation data is necessary, the prediction unit 23 outputs a request signal through communication to request compensation data.

FIG. 3 is a diagram illustrating an example of creating a received radio wave quality data string in the recording unit 21. The recording unit 21 records the received radio wave quality R at regular intervals, that is, every unit time (sampling time) t, and the data of the received radio wave quality R is recorded in the buffer of the recording unit 21. When the prediction unit 23 predicts the received radio wave quality, it uses n pieces of past data (n is an integer of 1 or more), defines a prediction target variable p (p is an integer of 1 or more), and calculates p×t from time T. When predicting the quality of received radio waves later, or when predicting whether to request compensation data after p×t, the recording unit 21 holds at least n pieces of data.

FIG. 3(a) shows that at a certain time T, data from R ₁ to R _n (R _n is the latest received radio wave quality data recorded at time T) is stored in the buffer of the recording unit 21. It shows the situation. At this time, the recording unit 21 outputs the received radio wave quality data string consisting of n pieces of data R ₁ to R _n as one data set to the learning data generating unit 22 and the predicting unit 23 .

FIG. 3(b) shows the state of the buffer in the recording unit 21 at time T+t when the received radio wave quality R is recorded next. Compared to the state at time T, data R _n+1 at time T+t is added, the oldest data R ₁ is deleted, and R ₂ to R _n+1 are stored in the buffer of the recording unit 21. . At this time, the recording unit 21 outputs a received radio wave quality data string consisting of n pieces of data from R ₂ to R _n+1 to the learning data generating unit 22 and the predicting unit 23 . After this, at time T+2t, the recording unit 21 outputs the received radio wave quality data strings from R ₃ to R _n+2 to the learning data generation unit 22 and the prediction unit 23. Thereafter, data is output in the same way.

FIG. 4 shows an example of the configuration of the learning data generation section 22. The learning data generation section 22 includes a reception radio wave quality data string storage section 221, a data effective range determination section 222, and a learning data generation section 223.

The storage unit 221 stores the received radio wave quality data string input from the recording unit 21. Further, the stored data string is output to the generation unit 223. Here, the received radio wave quality data string is R _xy [x is the data acquisition time, and y is the number (1≦y≦n) in the data string. ]. Taking the output from the recording unit 21 in FIG. 3(a) as an example, R _T1 =R ₁ , R _T2 =R ₂ , . . . R _Tn =R _n .

The determination unit 222 determines the effective range as learning data from the received radio wave quality data string input from the recording unit 21 . When training a recurrent neural network, rather than using time-series data in the normal state as training data, data that includes changes below the threshold _Rth of received radio wave quality R at which broadcast waves cannot be received is used as training data. It has been experimentally found that using this method is more effective in predicting deterioration of reception conditions. Therefore, the determination unit 222 determines the range to be used as learning data based on R _xn , which is the latest received radio wave quality data, so as to include the time period in which the reception condition deteriorated. The operation algorithm of the determination unit 222 will be described later. The determination unit 222 outputs a valid range (time information S, E at the beginning and end of the valid range) as learning data.

The generation unit 223 uses the time information S and E input from the determination unit 222 and the received radio wave quality data string R _xy stored in the storage unit 221 to generate a received radio wave quality data _string R ≦E, 1≦y≦n), and generate learning data using the received radio wave quality data string in this effective range. The operation of the generation unit 223 will be described later.

FIG. 5 is an example of a flowchart showing an operation algorithm of the determination unit 222. Each step will be explained below.

Step S1: Set Flag to False (Flag←False, hereinafter described similarly) and start the algorithm. Note that the algorithm executes a loop of steps S2 to S10 once for each data sampling time t.

Step S2: When the normal received radio wave quality is R _normal , and the received radio wave quality data string input at a certain time i is R _iy (1≦y≦n), compare the latest data R _in and R _normal. do. When R _in >R _normal , the process proceeds to step S3, and when R _in ≦R _normal , the process proceeds to step S8. In the flowchart, symbols i, a, b, S, E, etc. indicating time are symbols indicating a time number obtained by dividing the actual time by unit time t.

Step S3: Determine Flag. When Flag=True, the process proceeds to step S4, and when Flag=False, the process proceeds to step S7. (A state of True means that the time period can be training data.)

Steps S4 to S6: When R _in > R _normal and Flag=True, set b←i (S4). After that, S←a−M _s and E←b+M _e are set, and the values of S and E are output (S5). Here, M _s and M _e are margins, and are arbitrary values in the range of M _s ≧n+p and M _e ≧p. Then, the flag is returned to False (S6), the process is ended, and the process proceeds to step S10. Regarding the margin range, M _e is to secure p data sets ahead for predicting after p×t (p is the prediction target variable), and M _s is to secure p data sets for predicting after p × t (p is the prediction target variable), and M This is because n pieces of data are required.

Step S7: When R _in > R _normal and Flag=False, set a←i and proceed to step S10.

Steps S8 to S9: If R _in ≦R _normal in step S2, R _in is compared with a reception quality threshold R _th at which broadcast waves cannot be received (S8). If R _in <R _th , Flag←True is set (S9), and if R _in ≧R _th , the process is directly ended and the process proceeds to step S10.

Step S10: After completing the processing of each route, increment i (i←i+1) and return to the beginning of the loop. The algorithm restarts at step S2 at the next sampling instant (i+1).

How to determine the time information S, E of the effective range of learning data using the algorithm of FIG. 5 will be explained based on FIG. 6. In FIG. 6, the horizontal axis is time and the vertical axis is the received radio wave quality R _in (the latest received radio wave quality data in the received radio wave quality data string), and the received radio wave quality R _in (shown by the solid line) increases with time. It shows that it is changing.

Let R _th be the reception quality threshold at which broadcast waves become unreceivable. Let a and b be the times when the received radio wave quality of the broadcast wave is R _normal before and after the point where it falls below the threshold value R _th (Flag←True) (a and b are time numbers, so The actual time is a×t, b×t (t is unit time). In order to predict the future p×t using n pieces of past data, the margin parameters M _s ≧n+p and M _e ≧p are set. The time M _s (M _s ×t) earlier than time a is S (the start of the effective range), and the time M _e (M _e ×t) later than time b is E (the end of the effective range). time). A received radio wave quality data string between S and E (actual times S×t, E×t) is used as learning data.

That is, the first point in time (Flag←True) when the latest received radio wave quality data R _in becomes unreceivable (Flag←True) and the received radio wave quality at normal times becomes degraded ( a) and a second time point (b) at which the degraded received radio wave quality returns to the normal received radio wave quality, and a predetermined margin ( M _s , M _e ) are added to form a valid range (S to E) as learning data.

FIG. 7 is a diagram illustrating an example of creating learning data by the generation unit 223. FIG. 7(a) shows the received radio wave quality data string R _xy (S≦x≦E, 1≦y≦n) obtained from the storage unit 221, and based on this data string, the data shown in FIG. 7(b) is shown. Generate training data.

In FIG. 7B, R _xy (S≦x≦(Ep), 1≦y≦n) is prepared as input data. As the correct received radio wave quality (received radio wave quality after p×t), let R _(x+p)n (S≦x≦(Ep)) correspond to the input data R _xy (S≦x≦(Ep)), respectively. group. Furthermore, the determination as to whether compensation data should be requested after p×t can be set as a set of correct compensation request determinations, each corresponding to the input data. The reception quality at which the broadcast wave becomes unreceivable is R _th , and if R _(x+p)n <R _th , the correct compensation request determination is True; otherwise, it is False. At least one of the correct received radio wave quality and the correct compensation request determination is combined with input data R _xy (S≦x≦(Ep)), and the generated data set is output as a learning data set.

FIG. 8 shows an example of the configuration of the prediction unit 23 of the communication compensation requesting device 20. The prediction unit 23 includes an input data selector 231, a recurrent neural network (RNN) unit 232, and a compensation data request unit 233.

The input data selector 231 selects the received radio wave quality data string inputted from the recording section 21 or the learning data inputted from the learning data generation section 22, and outputs the selected data to the RNN section. In the learning mode, the input from the learning data generation section 22 is selected, and in the prediction mode, the input from the recording section 21 is selected.

The RNN unit 232 applies a model such as a simple RNN or LSTM (Long Short-Term Memory) as a recurrent neural network. When in learning mode, it learns based on learning data and adjusts internal parameters. In the prediction mode, the received radio wave quality data string from the recording unit 21 is input data, and the received radio wave quality after p x t (predicted received radio wave quality RP) or the judgment whether compensation data should be requested after p x t (predicted compensation) is performed. The request judgment JP) is output. Note that, as in another embodiment described later, the RNN unit may obtain a learned RNN model from the external learning server 110 and reflect this in the RNN unit, without performing the learning mode.

The compensation data requesting unit 233 is configured to request compensation data according to the predicted received radio wave quality RP or predicted compensation request determination JP input from the RNN unit 232 and the current received radio wave quality input from the recording unit 21. Outputs a request signal.

FIG. 9 is an example of a flowchart showing an operation algorithm of the compensation data requesting section 233. FIG. 9 shows an operation algorithm of the compensation data requesting section 233 when the RNN section 232 predicts the received radio wave quality after p×t. Each step will be explained below.

When the received radio wave quality data strings R ₁ to R _n are input from the recording unit 21, that is, at every time t, the flowchart is executed.

Step S11: Check the latest (current) received radio wave quality data R _n of the received radio wave quality data string, and determine the current received radio wave quality R _n and the reception quality threshold R _th at which broadcast waves cannot be received. compare. If the quality of radio waves is currently deteriorating prior to prediction, compensation data is requested immediately. Therefore, if R _n <R _th , the process proceeds to step S13; otherwise, the process proceeds to step S12, where it is determined whether compensation data should be requested based on the prediction.

Step S12: In the prediction mode, the RNN unit 232 predicts the received radio wave quality after p×t, and the RNN unit 232 outputs the predicted received radio wave quality RP. The compensation data requesting unit 233 prepares a threshold value RP _th for the predicted received radio wave quality, and compares the predicted received radio wave quality RP and the threshold value RP _th . If RP<RP _th , the process proceeds to step S13 (requesting compensation data); otherwise, the process proceeds to step S14.

Step S13: Output a request signal requesting compensation data. This request signal is transmitted to the compensation data server 120 via communication. After that, the process ends.

Step S14: If RP≧RP _th , no compensation data is requested and the process ends.

FIG. 10 is another example of a flowchart showing the operation algorithm of the compensation data requesting unit 233. FIG. 10 shows an operation algorithm of the compensation data requesting unit 233 when the RNN unit 232 determines whether to request compensation data after p×t. Each step will be explained below.

The flowchart is executed every time t when received radio wave quality data strings R ₁ to R _n are input from the recording unit 21.

Step S21: Check the latest (current) received radio wave quality data R _n of the received radio wave quality data string, and determine the current received radio wave quality R _n and the reception quality threshold R _th at which broadcast waves cannot be received. compare. Similarly to FIG. 9, if R _n <R _th , the process proceeds to step S23; otherwise, the process proceeds to step S22, where it is determined whether compensation data should be requested based on the prediction.

Step S22: When the RNN unit 232 determines whether to request compensation data after p×t, the RNN unit 232 sends a predicted compensation request determination JP (True if it is better to request compensation data, True if it is not necessary to request it). False) is output. It is determined whether the output from the RNN unit 232 is True. If True, the process proceeds to step S23 (request compensation data); otherwise, the process proceeds to step S24.

Step S23: Output a request signal requesting compensation data. This request signal is transmitted to the compensation data server 120 via communication. After that, the process ends.

Step S24: If the predicted compensation request determination JP is False, no compensation data is requested and the process ends.

As described above, the communication compensation requesting device 20 of FIG. be able to.

FIG. 11 shows another example of the configuration of the communication compensation requesting device. The communication compensation requesting device 20' includes a recording section 21 and a prediction section 23. In this embodiment, a recurrent neural network (RNN) is trained using an external learning server 110.

The recording unit 21 receives broadcast waves from receiving antennas 61 and 62, analyzes the broadcast waves, and records received radio wave quality R such as received electric field value, MER, and C/N. The configuration and operation of the recording unit 21 are basically the same as the recording unit 21 of the communication compensation requesting device 20 in FIG. 2, and generates the received radio wave quality data string in FIG. 3 at regular intervals (sampling time t). In this embodiment, the recording unit 21 outputs the received radio wave quality data string to the learning data generation unit 22 and transmits the output received radio wave quality data string to the external learning server 110.

The prediction unit 23 includes a recurrent neural network (RNN) therein, receives the received radio wave quality data string as input, and predicts the received radio wave quality after a predetermined period of time. However, in this embodiment, learning data generation and learning itself are performed by an external learning server 110. For example, the learning server 110 receives a received radio wave quality data string, generates therein the learning data shown in FIG. 7, trains a recurrent neural network (RNN) inside the server, and creates a trained RNN model. do.

After that, the prediction unit 23 receives the learned RNN model from the learning server 110 and reflects it in the recurrent neural network (RNN) inside the prediction unit 23 . After reflection, the received radio wave quality data string input from the recording unit 21 is used to predict the reception status (received radio wave quality) after a predetermined time, or to predict whether to request compensation data after a predetermined time. Then, based on the prediction, a request signal for the compensation data is output. That is, the operation of the prediction unit 23 is the same as the operation of the prediction unit 23 in the prediction mode of FIG. 2.

The communication compensation requesting device 20' of FIG. 11 does not require the learning data generating section 22 in the communication compensation requesting device, and has the effect of simplifying the configuration of the device.

As an example, a process of predicting the deterioration of reception conditions from actual measured data of beacon radio waves in the 21 GHz band, applying the algorithm shown in FIG. 9, and requesting compensation data was verified. Learning data was generated from the measured data of the reception situation (received electric field strength) on August 3, 2018, and the attenuation of the reception situation on September 4, 2018 was calculated using the model obtained as a result of learning based on the learning data. I predicted it.

FIG. 12 shows the beacon radio wave reception situation in the 21 GHz band on August 3, 2018. The horizontal axis is time (date and time), and the vertical axis is the received electric field [dBm] of the radio wave. The unit time (sampling period) is t = 1 second, and the received electric field is recorded every second. Figure 12 shows the situation from 16:48:0 to 19:12:0 on August 3. . It can be seen that there was rainfall during this period, and the attenuation started around 17:00 and stopped around 19:00.

Since there are no detailed regulations regarding satellite broadcasting in the 21 GHz band, there _is no definition regarding the reception quality at which broadcasting cannot be received. Assume that _th = -60 dBm and reception becomes impossible if the received electric field is attenuated by 10 dBm. In addition, with the aim of requesting compensation data 30 seconds before the reception situation worsens, by setting the number of input data n = 20 and the prediction target variable p = 30, the received electric field value for n × t = 20 seconds is calculated. Using this method, the user learns to predict the predicted received electric field value RP after p×t=30 seconds. To implement the prediction unit 23, TensorFlow/Keras Simple RNN (Non-Patent Document 3) was used.

FIG. 13 shows the range of reception status data determined as the effective range of learning data. The reception status data in FIG. 13 is the same as the reception status data in FIG. 12, but the time axis has been expanded. According to the algorithm of the determination unit 222 in FIG. 5, a = 17:42:51 and b = 18:18:17. In this example, M _s = 171, M _e = 102, and S = 17:40. It is determined that E=18:19:59 and the effective range of the learning data is set.

The RNN unit 232 of the prediction unit 23 performs learning in learning mode using data from 17:40:0 to 18:19:59 on August 3, 2018. During learning, the data was preprocessed, the average was set to -60 dBm, the standard deviation was set to 10, and standardized data was used. Essentially, data on the date, time, and time zone determined by the algorithm of FIG. 5 on any day may be used as the learning data, not limited to the above-mentioned time zone. However, in this example, the data was simplified for easy understanding, and the learning data was limited to only the period from 17:40:0 to 18:19:59 on August 3, 2018, and learning was performed.

To evaluate the prediction results, we used the beacon radio wave reception status in the 21 GHz band on September 4, 2018. FIG. 14 shows the reception status from 7:00:0 to 8:00:0 on September 4, 2018. The horizontal axis is time (date and time), and the vertical axis is the received electric field [dBm] of the radio wave. The period during which the actual received electric field was below the threshold received electric field R _th =-60 dBm was from 7:20:20 to 7:34:25. As can be seen from FIG. 14, since there is scintillation in the received electric field, if compensation data is requested only during the period from 30 seconds before falling below the threshold to the period during which it is below the threshold, a short gap may occur in the request period. Although scintillation may be removed from the received electric field, in this example, for simplicity, compensation data is requested for the period from 30 seconds before falling below the threshold to 30 seconds after the period falling below the threshold. period. Therefore, the ideal compensation data request period is from 7:19:50 to 7:34:55. In the verification according to this embodiment, this ideal compensation data request period was used as the standard for the compensation data request period.

First, as a conventional technique, consider a case where compensation data is requested simply by using a threshold value. When requesting compensation data, it was decided that 30 seconds of compensation data would be requested. When requesting compensation data is started when the received electric field falls below the threshold received electric field R _th =-60 dBm, the period for requesting compensation data is from 7:20:20 to 7:34:55. In this reception situation, since it would ideally have been necessary to request compensation data from 7:19:50, compensation data would be missing for the first 30 seconds.

Next, a case will be described in which compensation data is requested based on prediction of reception conditions according to the present invention. FIG. 15 shows changes in predicted received radio wave quality (predicted received electric field in the case of this embodiment) and the period for requesting compensation data. The solid line graph shows the actual change in the received electric field, and the dotted line graph shows the predicted change in the received electric field. The threshold value RP _th = R _th = -60 dBm of predicted received radio wave quality at which reception becomes impossible was set, and when requesting compensation data, 30 seconds of compensation data was requested in accordance with the algorithm shown in FIG. According to the reception situation prediction based on the present invention, the compensation data request period was from 7:19:56 to 7:35:6.

When compared with the ideal compensation data request period, compensation data is lost for 6 seconds from 7:19:50 to 7:19:55. On the other hand, compensation data is requested for 11 seconds from 7:34:56 to 7:35:6, even though it is not originally necessary. In this example, compensation data was requested even during a period that was not originally necessary, but compared to the case where the compensation data was lost for 30 seconds when using the threshold, only 6 seconds of compensation data was lost, so the period during which the data is lost is was reduced by 80%. As described above, according to the present invention, the period during which compensation data is lost can be significantly reduced compared to the conventional technology.

Next, a case will be described in which a margin is set for the threshold value of received radio wave quality. One existing proposal is to provide a margin for the threshold in order to request compensation data before reception becomes impossible. However, there is no proposal that mentions how to take the margin, so first we will consider the case where the threshold margin is set to include the entire ideal compensation data request period.

FIG. 16 shows the compensation data request period when a margin is provided for the threshold in the existing compensation data request method using the threshold. If the threshold with a margin is R _thmargin = -58 dBm and compensation data is requested for 30 seconds when the received electric field is below the threshold R _thmargin , the period from 7:19:10 to 7:35:26 will be compensated. The data request period has arrived. In this case, there will be no loss of compensation data compared to the ideal compensation data request period (7:19:50 to 7:34:55), but from 7:19:10 to 7:19:49. Compensation data will be requested for a period that is not originally necessary, for a total of 71 seconds, 40 seconds from 7:34:56 to 7:35:26. Note that the threshold value R _thmargin = -58 dBm with a margin was obtained by changing the threshold value from -60 dBm in predetermined increments (here, in 1 dBm increments) and searching for a threshold value that includes the entire request period of ideal compensation data. The minimum unnecessary compensation data request among the threshold values is determined as the threshold value with the least number of unnecessary compensation data requests.

Next, in the method of predicting received radio wave quality of the present invention, a case will be considered in which a threshold margin is set and the threshold value RP _th of predicted received radio wave quality is adjusted so as to include the entire request period of ideal compensation data. do.

FIG. 17 shows the compensation data request period based on the prediction of the reception situation when the threshold value is adjusted by providing a margin for the threshold value RP _th . The solid line graph shows the actual change in the received electric field, and the dotted line graph shows the predicted change in the received electric field. If the adjusted predicted received radio wave quality threshold is RP _th = -59 dBm and compensation data is requested when the predicted received radio wave quality after 30 seconds is below the threshold, the compensation data request period is from 7:19:05 to 7:00. This is 35 minutes and 16 seconds, which includes the entire request period for ideal compensation data. However, in this case, the 45 seconds from 7:19:05 to 7:19:49 and the 21 seconds from 7:34:56 to 7:35:16, a total of 66 seconds, are periods that are not originally necessary. Compensation data will be requested. The threshold adjusted RP _th = -59 dBm is similar to R _thmargin = -58 dBm, the threshold is changed from -60 dBm in steps of the predetermined received radio wave quality (in 1 dBm steps here), and the ideal compensation data request period is A threshold value that includes all of the threshold values is searched, and the threshold value that is the smallest among the searched threshold values and that requires the least amount of unnecessary compensation data is found. In this embodiment, a threshold value with an appropriate margin was found by searching, but an appropriate threshold value adjustment may be performed using a recurrent neural network.

When using the prediction results and threshold adjustment according to the present invention, compared to the conventional method of setting a margin on the threshold and adjusting the threshold so that the compensation data request period includes all the ideal compensation data request period. The system was able to reduce the originally unnecessary compensation data request period by about 7% from 71 seconds to 66 seconds.

In the above embodiment, the configuration and operation of the receiving system 100 and the communication compensation requesting device 20 have been described, but the present invention is not limited to this, and may be configured as a method of compensating for deterioration of reception conditions through communication. good. That is, the present invention may be configured as a method of predicting time-series changes in the reception status of broadcast waves and requesting compensation data based on the prediction results.

Note that a computer can be suitably used to function as the communication compensation requesting device 20 described above, and such a computer can run a program that describes processing contents for realizing each function of the communication compensation requesting device 20. This can be realized by storing the program in a storage unit and having the CPU of the computer read and execute the program. Note that this program can be recorded on a computer-readable recording medium.

Although the present invention has been described based on the drawings and examples, it should be noted that those skilled in the art will be able to easily make various changes and modifications based on the present disclosure. Therefore, the present invention should not be construed as being limited to the above-described embodiments, and various modifications and changes can be made without departing from the scope of the claims. For example, the functions included in each block, each step, etc. can be rearranged to avoid logical contradictions, and multiple constituent blocks and steps can be combined into one or divided. be.

10 Broadcast wave receiving device 20 Communication compensation requesting device 21 Recording unit 22 Learning data generating unit 23 Predicting unit 30 Compensation data receiving device 40 Broadcast wave compensation data switch 50 Reproducing device 61 Satellite antenna 62 Terrestrial antenna 100 Receiving system 110 Learning server 120 Compensation data server 221 Storage unit 222 Determination unit 223 Generation unit 231 Input selection unit 232 RNN unit 233 Compensation data request unit

Claims (5)

A communication compensation requesting device used in a receiving system that receives compensation data through communication when broadcast waves are attenuated,
a recording unit that records the received radio wave quality of the received broadcast wave and outputs a received radio wave quality data string;
comprising a recurrent neural network (RNN), using the received radio wave quality data string as input, predicting the received radio wave quality after a predetermined time, or predicting a determination as to whether to request the compensation data after a predetermined time; a prediction unit that outputs a request signal for the compensation data based on the prediction;
a learning data generation unit that generates learning data from the received radio wave quality data string;
Equipped with
In a learning mode, the recurrent neural network performs learning to predict received radio wave quality after a predetermined time or a determination as to whether to request compensation data after a predetermined time based on the learning data;
The learning data generation unit includes:
a storage unit that stores the received radio wave quality data string;
a determination unit that uses the received radio wave quality data string to determine a range that includes received radio wave quality that makes it impossible to receive broadcast waves and is effective as learning data;
a generation unit that selects the received radio wave quality data string in the range determined by the determination unit from the received radio wave quality data string in the storage unit and generates learning data;
The learning data includes a received radio wave quality data string that is input data, a correct received radio wave quality that is the received radio wave quality after a predetermined time from the time when the received radio wave quality data string is input, and the received radio wave quality data string. At least one of the correct compensation request determinations that determine whether compensation data should be requested after a predetermined time from the time when the compensation data is requested is set as a set,
The determination unit includes:
Check the latest received radio wave quality data in the above received radio wave quality data string, and check if the received radio wave quality during normal operation has deteriorated, with the latest received radio wave quality data being the received radio wave quality that becomes unreceivable. A first time point at which the radio wave quality becomes good and a second time point at which the degraded received radio wave quality returns to the normal received radio wave quality are specified, and a range between the first time point and the second time point is predetermined. A communication compensation requesting device , characterized in that a margin is added to the effective range of the learning data . The communication compensation requesting device according to claim 1 ,
The recording unit records the received radio wave quality at regular intervals, and includes a buffer for holding a predetermined number of received radio wave quality data, and records the received radio wave quality data held in the buffer at each fixed interval as received radio wave quality data. A communication compensation requesting device characterized in that it outputs data as a column. The communication compensation requesting device according to claim 1 or 2 ,
The prediction unit includes the recurrent neural network therein, receives the received radio wave quality data string as input, predicts the received radio wave quality after a predetermined time, or determines whether to request the compensation data after a predetermined time. an RNN unit that predicts and outputs the prediction;
The compensation data is sent to the compensation data server based on the latest input data of the received radio wave quality data string and the prediction of the received radio wave quality output from the RNN unit or the prediction of whether to request the compensation data. 1. A communication compensation requesting device, comprising: a compensation data requesting section that outputs a request signal. A program for causing a computer to function as the communication compensation requesting device according to any one of claims 1 to 3 . a broadcast wave receiving device that receives, demodulates and decodes broadcast waves;
4. The communication compensation requesting device according to claim 1, wherein a recurrent neural network (RNN) performs broadcasting after a predetermined time of reception based on time-series data of reception status of the broadcast waves. a communication compensation requesting device that predicts a wave reception condition and requests compensation data through communication when a deterioration of the reception condition is predicted;
a compensation data receiving device that receives and decodes the compensation data obtained in response to a request from the communication compensation requesting device;
a broadcast wave compensation data switch that switches and outputs the broadcast wave reception data output from the broadcast wave reception device and the compensation data output from the compensation data reception device based on the reception status of the broadcast waves;
A reception system equipped with.

Images