JP2018133745A - Radio reception method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To calculate by Q learning a threshold to be used to determine a superposition band and a non-superposition band.SOLUTION: A radio reception method comprises: receiving a reception signal having a desired signal and an interference signal superposed; determining electric power of an unnecessary signal using a predetermined threshold; classifying a subcarrier of the reception signal by a superposition band and a non-superposition band; performing likelihood estimation on desired signal electric power and noise electric power from a data subcarrier of the non-superposition band; and performing likelihood estimation on interference signal electric power from a data subcarrier of the superposition band. The radio reception method has a Q table classified by combinations of an electric power ratio DUR of the desired signal and interference signal and increase/decrease values of a plurality of different thresholds, and also comprises: calculating a DUR using a predetermined tentative threshold; increasing or decreasing a threshold corresponding to a maximum element among values of respective elements of the Q table corresponding to the DUR; and determining a threshold using Q learning in which values of respective corresponding elements of the Q table are recalculated and updated based upon a compensation function using the threshold.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a radio reception method in a radio communication system that performs multicarrier radio communication.

  In recent years, the depletion of frequency resources has become a problem due to the widespread use of various wireless communication systems, and studies on superposition transmission techniques that improve frequency utilization efficiency by sharing frequencies with a plurality of wireless signals are being promoted.

FIG. 4 shows an example of a wireless communication system sharing two frequency channels.
In FIG. 4, the wireless communication system includes wireless LAN base stations 51 and 52 and a wireless terminal 53. The wireless LAN base station 51 communicates using the frequency channel CH1 having the center frequency fa. The wireless LAN base station 52 communicates using the frequency channel CH5 having the center frequency fb (fa <fb). The wireless terminal 53 is arranged at a position where the wireless signals of both the wireless LAN base stations 51 and 52 reach, and receives a signal in which two wireless signals having center frequencies fa and fb partially interfere with each other. As another example of sharing the frequency band, there may be a case where systems of different communication methods such as a combination of a wireless LAN system and bluetooth (registered trademark) or WiMAX (registered trademark) share the frequency.

  In general, when such an interference signal exists, the communication characteristics are significantly deteriorated. However, on the assumption that the transmission method of the desired signal includes multicarrier and error correction coding, FEC (forward) is performed while suppressing the influence of interference. There is a technique for decoding and realizing accurate transmission (Non-Patent Document 1). This technique involves filtering the frequency components of the received signal in the received signal before the demodulation of the desired signal in the RF stage or IF stage or weighting the likelihood for the corresponding frequency component in the baseband region. Further, the present invention is characterized in that the influence of the interference signal is suppressed and demodulated and decoded.

  In order to implement such reception processing, a technique for detecting a frequency band (interference probability of each subcarrier) in which an interference signal exists and a desired signal-to-interference signal power ratio has been proposed (Non-Patent Document 2). In this technique, after receiving a radio frame, an interference band is repeatedly estimated within one frame using pilot symbols, and unnecessary signal power is estimated using the obtained interference band. FIG. 5 shows a configuration example of a conventional radio receiving apparatus using the technology.

  In FIG. 5, a reception signal in which an interference signal and noise power are superimposed on a desired signal is input to the radio reception apparatus. After the synchronization of the received signal is established, the cyclic prefix removal circuit 11 removes the cyclic prefix of the OFDM signal, the FTT circuit 12 converts the signal to a frequency domain signal, extracts the subcarrier component, and receives the received signal buffer 13 and the LLR calculation. Input to the circuit 14. Thereafter, processing is performed in units of subcarriers.

  The unnecessary signal power calculation circuit 21 generates a desired signal (replica signal) generated by the error correction coding circuit 16, the PSK / QAM mapping circuit 17, and the transmission path weight circuit 18 from the reception signal stored in the reception signal buffer 13. By subtracting, the unnecessary signal power is calculated and input to the superimposed band determination circuit 22. The superimposition band determination circuit 22 determines the unnecessary signal power using an appropriate threshold, determines that the unnecessary signal power exceeds the threshold, determines the superimposition band, and if the unnecessary signal power falls below the threshold, determines the non-superimposition band. The interference power is estimated at, the noise power is estimated at the subcarrier determined to be a non-superimposed band, and each estimated value is given to the LLR calculation circuit.

  The LLR calculation circuit 14 calculates a log-likelihood ratio (LLR) using the estimated values of interference power and noise power. At this time, the LLR is calculated in consideration of noise power and interference power in the superimposed band, the LLR is calculated in consideration of only noise power in the non-superimposed band, and the LLR is input to the decoding circuit 15 to obtain decoded received bits.

  The error correction coding circuit 16 and the PSK / QAM mapping circuit 17 generate a transmission replica signal based on the decoded received bits output from the decoding circuit 15, and the transmission replica weight and the transmission path estimation are performed by the transmission path weight circuit 18. The received replica signal (desired signal) for each subcarrier is generated by multiplying the value.

Masuno, Sugiyama, “Effect of frequency utilization improvement by multi-carrier superposition transmission,” IEICE Technical Report, vol.108, no.188, RCS2008-67, pp.85-90, August 2008. Yohei Shibata, Naoken Yoda, Tomoaki Ohtsuki, Satoshi Masuno, Takatoshi Sugiyama, “A Study on Log Likelihood Settings in Multi-Carrier Superposition Transmission,” IEICE Broadcast Technology Society, February 19, 2015

  If the residual power after subtraction of the replica signal under a specific condition such as a low SNR value is used for the threshold value used for determining the superimposed band and the non-superimposed band, an error may occur in the initial estimated value of the threshold value. there were. The conventional method is effective when the environment such as the occurrence of interference can be predicted to some extent, but the detection accuracy of interference information may deteriorate in an environment where unpredictable interference occurs. For example, when the interference power is high, the detection rate decreases, and when the interference power greatly varies, there is a problem that the variation cannot be followed.

  In a wireless communication system that performs multi-carrier wireless communication in an environment in which unpredictable interference occurs, such as high interference power or large fluctuations in interference power, the present invention sets a threshold value used for determining a superimposed band and a non-superimposed band as Q. It is an object of the present invention to provide a radio reception method that can improve the detection accuracy of interference information by dynamically calculating by learning.

  The present invention provides a radio reception method for receiving a reception signal in which a desired signal of a data subcarrier transmitted using a multicarrier superimposed transmission scheme and an interference signal that interferes with the desired signal are superimposed. The step of calculating the power of the unnecessary signal obtained by subtracting the replica signal of the desired signal for each subcarrier, the power of the unnecessary signal is determined using a predetermined threshold, and the subcarrier of the received signal is overlapped and non-superposed And classifying the desired signal power and noise power from the data subcarriers in the non-superimposed band, and estimating the interference signal power from the data subcarriers in the superposed band. It has a Q table that is classified by the combination of the interference signal power ratio DUR and the increase / decrease values of a plurality of different threshold values, and calculates the DUR using a predetermined temporary threshold value. Then, the threshold value corresponding to the largest element among the values of each element of the Q table corresponding to the DUR is increased or decreased, and the value of each element of the corresponding Q table is calculated based on the reward function using the threshold value. The threshold value is determined using Q-learning to be recalculated / updated.

  In the wireless reception method of the present invention, the reward function for determining the value of each element of the Q table takes an absolute value by subtracting a threshold value from each average power of the subcarriers in the superimposed band and the non-superimposed band. This is the setting that takes the maximum value when the difference is minimum.

  In the wireless reception method of the present invention, when the threshold value is increased or decreased, the threshold value is increased or decreased for elements that do not become the maximum among the values of each element of the Q table with a predetermined probability.

  In the present invention, even in an environment where unpredictable interference occurs, such as when the interference power is high or the fluctuation of the interference power is large, the threshold value used for determining the superimposed band and the non-superimposed band is dynamically calculated by Q learning, thereby Information detection accuracy can be increased.

It is a flowchart which shows the process sequence which optimizes the threshold value in this invention. It is a figure which shows the simulation result which shows the effectiveness of this invention. It is a figure which shows the relationship between Eb / No and a detection rate in this invention. It is a figure which shows an example of the radio | wireless communications system which shares two frequency channels. It is a figure which shows the structural example of the conventional radio | wireless receiver.

  The present invention is characterized by using a reward function for optimizing a Q table value by Q learning in order to optimize a threshold value used for determination of a superposed band and a non-superimposed band.

Here, the Q table is expressed as follows using each element Q _SyAx of the combination of the state s and the action a.

FIG. 1 shows a processing procedure for optimizing a threshold in the present invention.
In FIG. 1, a power ratio DUR (Desired to Undesired power Ratio) between a desired signal and an interference signal is calculated, and the state s is determined. Next, the selection of the action types in the state s _{m (m} = 0~y), essentially selects the Q value up action _{a n (n = 0~x),} Q value at a predetermined probability ρ also select the action a _n other than the maximum. Action a _n increase in the threshold, hold, given in three ways reduced. The increase / decrease value of the threshold is an arbitrary fixed value or variable.

Here, let R _SB and R _NSB be the average power of the subcarriers in the superimposed band and the non-superimposed band, and T _rp be the threshold used for determining the superimposed band and the non-superimposed band. The present invention aims to optimize the threshold value T _rp , and its index S (T _rp ) is a value obtained by subtracting the threshold value T _rp from R _SB and R _NSB and subtracting again between the absolute values. Is defined as follows.
S (T _rp ) = | R _SB −T _rp | − | R _NSB −T _rp |

A reward function Rω _t (s, a) that takes a maximum value when S (T _rp ) becomes minimum is calculated as follows.
Rω _t (s, a) = γ (s, a) * | 1 / S (T _rp ) _t | * (1+ (S (T _rp ) _t−1 −S (T _rp ) _t ))
In this way, the optimal value of the threshold T _rp is calculated while updating the Q table evaluated based on the reward function that determines the functional utility of the action a selected in the state s.

Note that γ (s, a), which is a part of the reward function, is a reduction coefficient and is expressed as follows.
γ (s, a) = 1 * (0.7 ^{GT (s, a)} )
0.7 is a value that determines the relationship between convergence stability and convergence time, and other values may be used. GT (s, a) here is a matrix having the same size as the Q table.

  GT (s, a) is the number of times that the action a in the state s is continuously selected. As the number of selections increases, γ (s, a) becomes a smaller value, so the increase in the reward function decreases. Thus, the Q values of the Q table in the state s and the action a converge.

Further, S (T _rp ) _t−1 −S (T _rp ) _t is a term to be evaluated regarding the validity of the action for setting the threshold. If a threshold value is set in the wrong direction, this value takes a negative value. However, in order to prevent the reward function from decreasing more than necessary, the effect is reduced by adding a numerical value 1.

FIG. 2 shows simulation results showing the effectiveness of the present invention.
Here, the accuracy of the number of packets and the detection rate of the conventional method (Theoretical Threshold) and the present invention method (Q-Learning Threshold) in DUR = 0 [dB] and Eb / No = 0 [dB] are shown. As the number of packets increases, it can be seen that Q-learning is completed and detection is more accurate than the conventional method.

The parameters used for this simulation are as follows.
SNR range 0 to 10 dB
DUR -3, 0, 3dB
Number of subcarriers 62
Packet size 7 symbols (2 pilots, 5 data)
Number of packets 1430
Initial threshold value 0.8
Threshold change rate 0.05
Duplicate bandwidth 20/64 (31.25%)

FIG. 3 shows the relationship between Eb / No and detection rate in the present invention.
Here, it can be confirmed that the detection rate in the present invention is improved from the conventional method in the low SNR region, and is almost the same as that in the conventional method even in the high SNR region. Therefore, a threshold value for Eb / No may be set separately, and the conventional method and the present invention method may be switched according to the magnitude relationship between Eb / No and the threshold value.

DESCRIPTION OF SYMBOLS 11 Cyclic prefix removal circuit 12 FFT circuit 13 Reception signal buffer 14 LLR arithmetic circuit 15 Decoding circuit 16 Error correction encoding circuit 17 PSK / QAM mapping circuit 18 Transmission path weight circuit 21 Unnecessary signal power calculation circuit 22 Superimposition band judgment circuit

Claims (3)

In a radio reception method for receiving a reception signal in which a desired signal of a data subcarrier transmitted using a multicarrier superimposed transmission scheme and an interference signal that interferes with the desired signal are superimposed,
Calculating the power of the unnecessary signal obtained by subtracting the replica signal of the desired signal from the received signal for each subcarrier;
The power of the unnecessary signal is determined using a predetermined threshold, the subcarriers of the received signal are classified into a superimposed band and a non-superimposed band, and the desired signal power and noise power are estimated from the data subcarriers of the non-superimposed band. And the maximum likelihood estimation of the interference signal power from the data subcarriers in the superposed band,
It has a Q table divided according to the combination of the power ratio (DUR) of the desired signal and the interference signal and the increase / decrease values of a plurality of different thresholds, calculates the DUR using a predetermined temporary threshold, The threshold value corresponding to the largest element among the values of each element of the corresponding Q table is increased or decreased, and the value of each element of the corresponding Q table is recalculated and updated based on the reward function using this threshold value. The wireless reception method, wherein the threshold is determined using Q-learning. The wireless reception method according to claim 1,
The reward function for determining the value of each element of the Q table takes an absolute value by subtracting a threshold value from each average power of the subcarriers in the superimposed band and the non-superimposed band, and the difference between the absolute values is minimum. A wireless reception method characterized in that the maximum value is set when The wireless reception method according to claim 1,
The radio reception method according to claim 1, wherein when the threshold value is increased or decreased, the threshold value is increased or decreased with respect to an element that does not become the maximum among the values of each element of the Q table with a predetermined probability.

Images