JP5249541B2 - Propagation path estimation apparatus and reception apparatus - Google Patents

Description

  The present invention relates to a propagation path estimation apparatus and a reception apparatus, and can be applied to, for example, a multicarrier communication system that communicates training symbols that are known information followed by data symbols that are unknown information.

As described in Non-Patent Document 1, in the IEEE802.11n standard, a propagation path is estimated using a training symbol (preamble signal), and a data path following the training symbol is used to estimate a propagation path. Is specified to perform demodulation.
Supervised by Masahiro Morikura and Shuji Kubota, “Revised 802.11 High-Speed Wireless LAN Textbook”, published by Impress Corporation in January 2005, Chapter 9.6 pp. 212-214,

  However, in the conventional propagation path estimation method, for example, when the moving speed of a mobile device used for communication is high, or when a reflective object (such as a human being) that is a component of the propagation path moves, The amplitude and phase fluctuate greatly due to fluctuations in the propagation path, the accuracy of propagation path estimation using the training symbol is lowered, and there is a problem that good communication (data demodulation) cannot be performed.

  Also, if there is an error between the local frequency at the transmitter and the local frequency at the receiver, this is the same as when phase fluctuations occur in the propagation path, and the accuracy of propagation path estimation using conventional training symbols is low. Therefore, there arises a problem that good communication (demodulation) cannot be performed.

  Therefore, a propagation path estimation apparatus, a propagation path estimation method, and a reception apparatus that can increase propagation path estimation accuracy and improve demodulation accuracy are desired.

The first of the present invention is a channel estimation apparatus provided subsequent to the training symbol is a known information to the receiving apparatus for receiving a signal having a data symbol is unknown information, (1) the received signal of the training symbol (2) By applying the estimated propagation path estimated value and reducing the influence of the propagation path characteristics from the received signal of the data symbol , by comparing with the threshold value, a hard decision value obtained Ru bit hard decision unit, and a follow-up means with its hard decision value and the known information of the data symbols, the channel tracking unit also continues at the reception signal period of the estimated data symbols of the channel It is characterized by having.

  A receiving apparatus according to a second aspect of the present invention includes the propagation path estimating apparatus according to the first aspect of the present invention.

  According to the present invention, even in a data symbol period in which unknown information is transmitted, by estimating a propagation path, when a moving speed of a corresponding mobile device is high or a reflection of a person or the like Even when the object moves or when the frequency error between the transmitter and the receiver is large, a more accurate propagation path estimation value can be obtained and good communication can be provided.

(A) Main Embodiment Hereinafter, an embodiment of a propagation path estimation apparatus and a reception apparatus according to the present invention will be described in detail with reference to the drawings.

(A-1) Overall Configuration and Operation of Propagation Channel Estimation Device FIG. 1 is a block diagram illustrating a configuration of a propagation channel estimation device 1 according to an embodiment. Note that the receiving apparatus of the embodiment uses the propagation path estimation apparatus of the embodiment.

  In FIG. 1, the propagation path estimation device 1 includes a bit hard decision unit 2, a propagation path provisional decision unit 3, a frequency domain repetition processing unit 4, a time domain processing unit 5, and a propagation path smoothing unit 6. Although omitted in FIG. 1, there is also provided control means for appropriately switching the operations of the units 2 to 6 depending on whether the symbol is a training symbol or a data symbol.

  The propagation path estimation apparatus 1 receives a reception signal converted into the frequency domain by FFT (Fast Fourier Transform).

  The bit hard decision unit 2 does not perform any processing when the input received signal is a known training symbol. If the input received signal is a data symbol, the bit hard decision unit 2 estimates the channel and performs propagation path smoothing. An unknown transmission bit is determined (hard determination) using the propagation path estimation value given from the conversion unit 6.

  The data symbols determined by the bit hard decision unit 2 are transmitted to the propagation path tentative determination unit 3, the frequency domain repetition processing unit 4, the time domain processing unit 5, and the propagation path smoothing unit 6 in the same manner as known training symbols. It is processed. Therefore, the functions of the propagation path tentative determination unit 3, the frequency domain repetition processing unit 4, the time domain processing unit 5, and the propagation path smoothing unit 6 will be described using an example in which the input received signal is a known training symbol. To do.

  The propagation path tentative determination unit 3 tentatively determines propagation path characteristics (hereinafter, the propagation path characteristics are appropriately abbreviated as propagation paths) using known training symbol information.

  The frequency domain repetitive processing unit 4 performs repetitive processing using an interpolation function on the provisionally determined propagation path, and interpolates specific subcarrier information.

  The time domain processing unit 5 converts the interpolated propagation path estimated value into the time domain by IFFT (Inverse Fast Fourier Transform) processing, and then removes unnecessary information such as noise and interference by applying a window function. To do.

  The propagation path smoothing unit 6 converts the output of the time domain processing unit 5 into the frequency domain by FFT processing, smoothes it with the past propagation path estimation value, and forms a propagation path estimation value to be output.

  When the received signal (the received signal converted into the frequency domain by FFT) is a known training symbol, the bit hard decision unit 2 does not function and is propagated from the known training symbol information by the propagation path provisional decision unit 3. The road is provisionally determined. The propagation path that has been provisionally determined is subjected to repetition processing using an interpolation function by the frequency domain repetition processing unit 4, and specific subcarrier information is interpolated. The interpolated propagation path estimated value is converted into the time domain by the IFFT process by the time domain processing unit 5 and then subjected to a window function to remove unnecessary information such as noise and interference. The output of the time domain processing unit 5 is converted into the frequency domain by the FFT processing by the propagation path smoothing unit 6, smoothed with the past propagation path estimation value, and then output.

  On the other hand, when the received signal is a data symbol, the bit hard decision unit 2 applies the previously estimated propagation path estimation value given from the propagation path smoothing unit 6 and makes a hard decision on the unknown transmission bit. The As a result, unknown transmission bits (data symbols) can be handled as known symbols in the propagation channel estimation operation, and the propagation channel temporary determination unit 3, the frequency domain repetition processing unit 4, the time domain processing unit 5, and the propagation channel The smoothing unit 6 performs the same processing as in the case of the training symbol, and estimates the propagation path.

  As described above, even in the data symbol period in which unknown information is transmitted, by estimating the propagation path, when the moving speed of the mobile device is fast, when a reflecting object such as a person moves, Even when the frequency error between the receivers is large, a more accurate propagation path estimation value can be obtained, and good communication (data demodulation) can be provided.

  Below, the detail of each part 2-6 is demonstrated about the case where it applies to the MIMO (Multi Input Multi Output) communication in IEEE802.11n standard.

(A-2-1) Bit hard decision unit 2
A subcarrier f, a transmission signal vector at time t (referred to as a vertical vector; hereinafter the same) is x (f, t), a channel matrix is H (f, t), and a noise vector is n (f, t). Then, the received signal vector y (f, t) can be expressed as in equation (1).

y (f, t) = H (f, t) x (f, t) + n (f, t) (1)
The bit hard decision unit 2 estimates a transmission signal vector x (f, t) using a technique such as ZF (Zero-Forcing), MMSE (Minimum Mean-Square Error), MLD (Maximum Likelihood Detection), and the like. judge. Here, the transmission signal vector x (f, t) is obtained by using the propagation channel matrix (propagation channel estimated value) ^ H (f, t−1) estimated by the propagation channel smoothing unit 6 one symbol before. Estimate (note that ^ represents the estimated value).

(A-2-2) Propagation path temporary determination unit 3
A transmission signal vector estimated in the bit hard decision unit 2 (in the case of a data symbol period) or known information transmitted in a training symbol (in the case of a training symbol period) is defined as x ′ (f, t). .

In the data symbol period, the temporary propagation path determination unit 3 receives a reception signal matrix Y (f, t) obtained by concatenating reception signal vectors y (f, t) corresponding to the number of transmission antennas in the column direction, and an estimated transmission signal vector. A transmission signal matrix X ′ (f, t) (or its inverse matrix X ′ ⁻¹ (f, t)) obtained by concatenating x ′ (f, t) in the column direction is generated, and the subcarrier f and time t The propagation path temporary determination value matrix H ′ (f, t) is obtained by the equation (2).

H ′ (f, t) = Y (f, t) X ′ ⁻¹ (f, t) (2)
In the training symbol period, the temporary propagation path determination unit 3 divides the received signal for each reception antenna by the known information, thereby estimating the propagation path estimated value (propagation path temporary determination) by combining the propagation paths from the plurality of transmission antennas. A value H ′ (f, t)) is obtained.

(A-2-3) Frequency domain repetition processing unit 4
In IEEE802.11n, when performing transmission using a plurality of transmission antennas, cyclic shift diversity (CSD) is employed in which the same training symbol is time-shifted for each transmission antenna and multiplexed and transmitted. Therefore, the receiver estimates the propagation path for each transmission antenna from the training symbols transmitted with time shift. In order to estimate the propagation path for each transmission antenna, after obtaining a propagation path estimated value in which a plurality of propagation paths are combined by the propagation path temporary determination unit 3 as described above, the propagation path estimated value is time-converted by IFFT. It is necessary to separate the delay profiles by converting into a combined delay profile of the region and extracting a time region necessary for each transmission antenna by a window function.

  However, in IEEE802.11n, a subcarrier (ineffective subcarrier) corresponding to a guard band (GB) and a direct current component (DC) does not transmit information. When the FFT size of the training symbol is different from the number of effective subcarriers, the obtained combined delay profile has a waveform distortion in which an impulse due to the propagation path estimation value of the ineffective subcarrier being 0 is spread in the time domain. At this time, if the delay profile is cut out by the time window, the information spread outside the window is discarded, and the estimation accuracy deteriorates.

  Therefore, in this embodiment, the waveform distortion in which the impulse spreads in the time domain is compensated by interpolating the propagation path of the subcarrier in which information is not transmitted in the frequency domain.

  It is assumed that there is a frequency domain propagation path estimation value H ′ (f, t) and some subcarrier estimation information is missing. This missing propagation path estimation value is interpolated by iterative processing using the interpolation function F (n).

Let N _rack be the number of subcarriers for which the channel estimation value is missing. The subcarrier index in which the channel estimation value is missing is m _k (k = 1, 2,..., N _rack ), the output of the channel preliminary determination unit 3 is H ′ (f, t), and the interpolation function F ( If the length of the n) of the n _F, is missing channel estimation value H '(m _{k, t)} is updated as (3).

By updating H ′ (m _k , t) according to equation (3) from k = 1 to N _rack , all missing propagation path estimation values are interpolated. Further, the update processing from k = 1 to N _{rack for} H ′ (m _k , t) is set as one iteration (hereinafter referred to as iteration), and interpolation is performed by repeatedly executing this iteration. Improve accuracy.

Figure 2 _is a flow chart showing the processing in the frequency domain iterative processing unit 4 including the _{N iterate} iterations. The parameter k represents the number of missing transmission path estimation values in one iteration, and the parameter j represents the number of iterations.

The parameter j is set to an initial value 1 and the first iteration is started (step 100). In the first iteration process, the parameter k is set to the initial value 1 (step 101), and after performing the interpolation shown in the equation (3) (step 102), it is confirmed that the parameter k does not exceed N _rack. (Step 103) The parameter k is incremented by 1 (Step 104), and the process returns to Step 102. When the parameter k exceeds N _rack , in other words, when the first iteration is completed, it is confirmed that the parameter j does not exceed N _iterate (step 105), and the parameter j is incremented by 1 (step 106). Return to step 101. Thereby, the second iteration is started.

Second _{to N iterate} th iteration is also performed as _above, the iteration process of the _{N iterate} th is completed, the series of processes ends. Through the processing as described above, an interpolated propagation path estimation value H ″ (f, t) is formed from the incomplete propagation path estimation value.

Here, as an interpolation function F (n) for interpolating the propagation path estimation value, for example, a DFT (Discrete Fourier Transform) function of some window function W (t) in the time domain is used. Therefore, the equation for deriving the interpolation function F (n) is as shown in equation (4). In the equation (4), A represents an arbitrary value, and N _FFT represents the FFT length. Hereinafter, a rectangular window will be described, but the same effect can be expected by applying other window functions. The window function W (t) in the time domain is a function having a value of 1 at a sample time where a delay path can exist and a value of 0 at other times as shown in the equation (5). In Equation (5), D represents a set of sample times in which delay paths can exist.

  In the data symbol period, the propagation path estimation by the iterative process is performed by including the propagation path estimation value determined as the determination error in the subcarrier index in which the propagation path estimation value is missing. The determination error can be determined by, for example, determining a square error between the previous channel estimation value and the current channel estimation value as a threshold. That is, when the channel estimation value of a subcarrier is determined to be a determination error, as described above, the channel estimation value of the subcarrier is updated in the same manner as the missing subcarrier.

  The frequency domain repetition process requires an initial value of a channel estimation value (including a channel estimation value determined to be a determination error) of a subcarrier index for which a channel estimation value is missing. One of the following initial values (a) to (d) is set as the initial value of the subcarrier to be interpolated when performing the iterative processing.

(A) 0 or any other fixed value (b) Propagation path estimated value for the previous time (c) Average of propagation path estimated values for the left and right subcarriers (d) In addition, averaging in time direction and frequency direction Propagation channel estimation values subjected to (simple average, weighted average, etc.) As the configuration of the frequency domain repetition processing unit 4, by adopting a configuration in which the intermediate result of the expression (3) is held in a memory, iterative processing 2 The amount of computation after the first iteration can be reduced. Hereinafter, this will be described with reference to FIGS. 3 and 4.

As an example, consider the case where FFT size = interpolation function length = 32 and the number of missing estimated values N _rack is 4. When a certain estimated value H (A) is interpolated, since the interpolation formula is a product-sum operation as shown in the above-described formula (3), the first iteration is 32 taps as shown in FIG. This is equivalent to the FIR filtering. Here, since the known estimated value is not updated, the output of the tap to which the known estimated value is input is the same in the next and subsequent iterations. Therefore, the sum of the outputs of the taps to which the known estimated values are input is held in the memories M1 and M2, thereby updating the estimated value H (A) after the second iteration as shown in FIG. Then, after performing a process equivalent to 4-tap FIR filtering, it may be combined with the contents stored in the memories M1 and M2. As described above, the calculation amount to which the interpolation function F (n) after the second iteration is applied can be reduced to N _rack / filter length NF of the first iteration.

  Furthermore, in the frequency domain repetitive processing unit 4, it is possible to reduce the amount of calculation using the feature of the interpolation function F (n).

The interpolation function F (n) has a tap having a value of 0 in the frequency domain when the function length N _F is a multiple of the length N _rect of the time domain rectangular filter. Since it is not necessary to calculate a tap having a value of 0, the amount of calculation can be reduced by devising the setting of the function length N _F and the rectangular filter length N _rect .

For example, when N _F = N _FFT = 64 and N _rect = 32, the interpolation function F (n) according to the above equation (4) is as shown in FIG. 5 (however, in FIG. 5, in equation (4) Unlike n, n is represented so as to change in a positive and negative range centered on 0), and there are 31 taps of 0. When this interpolation function F (n) is applied and the calculation of the tap of 0 is omitted, the calculation amount can be reduced to 33/64 (about 1/2) of the original calculation amount. In this example, the real element I other than 0 always has the same value other than F (0), and the imaginary element Q is symmetric about F (0). The amount of calculation can be further reduced.

Various interpolation functions F (n) can be applied. For example, a sinc function as shown in equation (6) can also be applied. Depending on the shape of the window function, the interpolation function represented by the equation (4) may be a sinc function. In formula (6), A and B are arbitrary values.

(A-2-4) Time domain processing unit 5
The time domain processing unit 5 performs processing in the time domain in order to separate the propagation path for each transmission antenna and remove noise and interference components. Expression (7) shows processing performed by the time domain processing unit 5.

  The time domain processing unit 5 converts the propagation path estimation value H ″ (f, t) obtained by the frequency domain repetition processing unit 4 into the time domain by IFFT, and multiplies the window function W (k) in the time domain. Thus, the propagation path for each transmission antenna is separated and the noise and interference components are removed simultaneously. Here, the rectangular window used in the frequency domain repetition processing unit 4 is used as the window function W (k). However, samples having a value of 1 and a value of 0 may be different, a Hamming window, Other window functions such as a Kaiser window or a Blackman window may be used. When separating multiplexed channels using CSD, it is necessary to use the window functions W (k) and shift that are time-shifted by the samples shifted by CSD. Thereafter, the time shift (shift) of the CSD is compensated (ROT {} in the equation (6) represents such processing), and is converted again into the frequency domain by FFT, so that H ′ ″ (f, t) Get.

  However, in an environment where the frequency domain repetitive processing unit 4 cannot sufficiently recover a subcarrier lacking a channel estimation value (such as a low SINR (signal-to-interference noise power ratio) environment), the characteristics are obtained by time domain processing. There is a possibility of deteriorating. Therefore, the time domain processing unit 5 does not perform the time domain processing, and the channel estimation value H ″ (f, t) obtained by the frequency domain repetition processing unit 4 is used as it is. It also has a function of turning off the time domain processing for making the output H ′ ″ (f, t) from For example, when the measurement result of the SINR measurement unit (not shown) is low, the time domain processing unit 5 turns off the time domain processing or turns off the time domain processing according to the operation of the corresponding operator by the user. .

(A-2-5) Propagation path smoothing unit 6
The propagation path smoothing unit 6 determines the propagation path estimation value H ′ ″ (f, t) obtained by the time domain processing unit 5 and the propagation path estimation value ^ H (f, t−1) one symbol before. Is used to obtain a channel estimation value ^ H (f, t) smoothed in the time direction by the equation (8). In the equation (8), α is a forgetting factor.

  In equation (8), the channel estimation value smoothed in the time direction by the forgetting average is obtained, but even if smoothing means such as interval average or moving average is used, when equation (8) is applied, Similar effects can be expected.

(A-3) Effect of Embodiment According to the above embodiment, since the propagation path is continuously estimated from the training symbol period even in the data symbol period, when the moving speed of the mobile device is fast, Even when a reflecting object such as a moving object or a frequency error between a transmitter and a receiver is large, an accurate propagation path estimation value can be obtained in a data symbol period and good communication (data demodulation) should be provided. Is possible.

  Hereinafter, effects (simulation results) when applied to IEEE802.11n will be described. FIG. 6 shows parameters of a specific example system whose effect has been confirmed. Two transmission antennas (Tx1, Tx2) are used, and a symbol obtained by shifting a transmission signal by 400 ns (8 samples) is transmitted from one transmission antenna Tx1 to the other transmission antenna Tx2.

  FIG. 7 shows the characteristics of the above-described specific example system. The horizontal axis represents the number of follow-ups (number of data OFDM symbols used for propagation path estimation / 2). In FIG. 7, the state in the training symbol period is omitted. The vertical axis represents the mean square error (MSE) of the propagation path estimated value.

  The hold in FIG. 7 is a case where the prior art is applied, and when the channel estimation value at the end of the training symbol period is held and applied to the data symbol period, the MSE gradually increases. On the other hand, in the embodiment, characteristic deterioration can be suppressed.

  From FIG. 7, it can be seen that the smaller the forgetting factor α, the better the MSE with time (number of follow-ups), and the more the iteration, the better the MSE. In particular, when the forgetting factor α is 0.2, the MSE is 97% at the position of the number of follow-ups 1000 (the difference in the number of iterations is irrelevant at the position of the number of follow-ups 1000) as compared with the prior art. It turns out that the above is also improved.

(B) Other Embodiments In the description of the above-described embodiment, various modified embodiments have been referred to. However, modified embodiments as exemplified below can be given.

  Although not mentioned in the description of the above embodiment, all or part of the bit hard decision unit 2, the propagation path provisional decision unit 3, the frequency domain repetition processing unit 4, the time domain processing unit 5, and the propagation path smoothing unit 6 are included. Alternatively, it may be configured with a dedicated IC chip, may be configured with hardware such as a discrete component, or may be realized with a DSP. Further, the CPU and the program executed by the CPU include all or part of the bit hard decision unit 2, the propagation path tentative determination unit 3, the frequency domain repetition processing unit 4, the time domain processing unit 5, and the propagation path smoothing unit 6. It may be realized.

  In FIG. 1, the propagation path estimation value of the propagation path estimation apparatus 1 is described as being given to a subsequent circuit (such as a data demodulation section), but the bit hard decision section 2 in the propagation path estimation apparatus 1 is used as a data demodulation section. It may be used as it is.

  In the above embodiment, the training symbol processing system and the data symbol processing system are partially shared. However, the training symbol processing system and the data symbol processing system are configured separately. You may make it do.

In the description of the above embodiment, the case where the present invention is applied to MIMO communication in IEEE802.11n has been described, but the present invention can also be applied to other wireless communication systems. That is, the present invention can be applied to other than MIMO communication, and is not limited to a multicarrier system. Further, the data modulation method and the like are not limited. For example, the present invention is also applied to WLAN, WPAN, WiMAX, MBWA, 3GPP, 3GPP2, and OFDM, OFDMA, SC-FDMA, SC-CDMA, MC-CDMA, etc., which are communication systems standardized in digital television broadcasting. Applicable.

It is a block diagram which shows the whole structure of the propagation path estimation apparatus which concerns on embodiment. It is a flowchart which shows the process in the frequency domain repetition process part of embodiment. It is explanatory drawing (1-1) of the reduction method of the amount of calculations in the frequency domain repetition process part of embodiment. It is explanatory drawing (1-2) of the reduction method of the amount of calculations in the frequency domain repetition process part of embodiment. It is explanatory drawing (2) of the reduction method of the calculation amount in the frequency domain repetition process part of embodiment. It is explanatory drawing of the parameter of the system applied in order to demonstrate the effect in the specific example system of embodiment. It is a graph which shows the effect in the example system in Drawing 6.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... Propagation path estimation apparatus, 2 ... Bit hard decision part, 3 ... Propagation path temporary determination part, 4 ... Frequency domain repetition process part, 5 ... Time domain process part, 6 ... Propagation path smoothing part

Claims (10)

In a propagation path estimation apparatus provided in a receiving apparatus that receives a signal having a data symbol that is unknown information following a training symbol that is known information,
Initial propagation path estimation means for estimating the propagation path using the received signal of the training symbol;
Applying the estimated channel estimation value, after reducing the influence of the propagation path characteristics from the received signal of the data symbol, and the resulting Ru-bit hard decision unit a hard decision value by comparison with the threshold value, the hard decision value data A channel estimation device comprising: a tracking unit having known information of a symbol, and a channel tracking unit that continues channel channel estimation even during a data symbol reception signal period. The propagation path estimation apparatus receives a signal in which information is distributed to a plurality of subcarriers,
The initial propagation path estimation means is
A first propagation path tentative determination unit for obtaining a propagation path tentative determination value from the input training symbol signal ;
From the first propagation path tentative determination unit, a propagation path tentative determination value is input, the propagation path tentative determination value is set as an initial value, and a predetermined interpolation function is determined for a specific subcarrier in the frequency domain. A frequency domain repetitive processing unit that performs repetitive interpolation processing with
A time domain processing unit that applies a predetermined window function in the time domain to the propagation path estimated value interpolated in the frequency domain output from the frequency domain iterative processing unit and can be switched ON / OFF of the processing. The propagation path estimation apparatus according to claim 1.   The propagation path estimation apparatus according to claim 2, wherein the frequency domain interpolation function applied by the frequency domain repetition processing unit is a sinc function.   The propagation path estimation apparatus according to claim 2, wherein the frequency domain interpolation function applied by the frequency domain repetition processing unit is a discrete Fourier transform of a window function predetermined in the time domain.   5. The window function in a predetermined time domain is a function having a value of 1 at a sample time in which a delay path can exist and a value of 0 at other times. Propagation path estimation apparatus.   6. The propagation path estimation apparatus according to claim 5, wherein the frequency domain repetition processing unit omits an interpolation calculation for a coefficient of an interpolation function having a value of 0 or a value close to 0.   The frequency domain repetitive processing unit holds calculation results other than the subcarriers to be interpolated, omits similar calculations in the subsequent repetitive calculations, and applies the held calculation results. Item 3. The propagation path estimation apparatus according to Item 2. The propagation path follower is
A second propagation path temporary estimation unit that temporarily estimates a propagation path using the output of the bit hard decision unit;
A propagation path interpolation unit having the same configuration as the frequency domain repetition processing unit and the time domain processing unit;
The propagation path estimation apparatus according to claim 1, further comprising: a propagation path smoothing unit that smoothes an output of the propagation path interpolation unit. 9. The propagation path estimation apparatus according to claim 8 , wherein the propagation path interpolation section also confirms whether or not there is an error in the hard decision result of the data symbol in the bit hard decision section. Receiving apparatus characterized by having a propagation path estimation apparatus according to any one of claims 1-9.

Images