CN1221096C

CN1221096C - Channel estimation method for orthogonal frequency-division multiplexing communication system

Info

Publication number: CN1221096C
Application number: CN 03118746
Authority: CN
Inventors: 段方庆; 刘应状; 谌璟; 朱光喜
Original assignee: Huazhong University of Science and Technology
Current assignee: Huazhong University of Science and Technology
Priority date: 2003-03-08
Filing date: 2003-03-08
Publication date: 2005-09-28
Anticipated expiration: 2023-03-08
Also published as: CN1437338A

Abstract

The present invention discloses a channel estimation method for an orthogonal frequency division multiplexing communication system. The present invention comprises the following steps that (1), an OFDM pilot signal for channel estimation is generated by a system transmitter, the step comprises a pilot insertion mode and pilot value selection, and the generated pilot signal is converted into a time-domain pilot signal through inverse Fourier conversion; (2), the time-domain pilot signal is transmitted on a wireless multi-path fading channel; (3), Fourier conversion is carried out for the time-domain pilot signal by the system receiver so as to obtain a frequency-domain pilot signal; channel estimation is carried out for a wireless channel; firstly, the estimation value of a sub channel, on which the pilot signal exists, is obtained through an LMMSE estimation or a simplified approximation method obtained from an MMSE estimation; then, the estimation value of the whole wireless channel is obtain by means of a filter interpolation method. The present invention is capable of effectively estimating the channel response value of an OFDM communication system and provides a feasible method for solving contradiction between improvement of estimated accuracy and reduction of calculation complexity in an existing channel estimation.

Description

Channel estimation method in orthogonal frequency division multiplexing communication system

Technical Field

The present invention relates to a channel estimation method in a mobile communication system, and more particularly, to a channel estimation method in an orthogonal frequency division multiplexing communication system.

Background

Currently, the development of mobile communication systems provides users with personal mobile communication terminals capable of supporting various new services, such as multimedia communication. Because of the large amount of data that each service requires to transmit, communication systems require the use of higher bit transmission rates. The use of higher bit rates in conventional single carrier systems can present difficulties in the efficient reception of signals due to inter-symbol interference (ISI) and deep frequency selective fading of the wireless channel.

One approach to solving these problems is to employ Orthogonal Frequency Division Multiplexing (OFDM) techniques in a wireless mobile environment. In an OFDM communication system, signals are transmitted and transmitted on a plurality of orthogonal carriers having a bandwidth less than the coherence bandwidth of the channel to overcome the problem of frequency selective fading of the signals caused by multipath fading channels. The ISI problem can be solved by using a guard interval. OFDM systems are currently used in europe for digital audio broadcasting and Digital Video Broadcasting (DVB) systems. It is also used in Asymmetric Digital Subscriber Lines (ADSL) to transmit high speed data.

A common channel estimation method is based on adaptive signal processing, where the channel is considered slowly varying. The channel parameters estimated at a particular time are dependent on the received data and channel parameters at the previous time. In the case of multipath fading and fast varying channels, such as in a high data rate OFDM mobile system, to overcome the adverse effects of the channel, pilot signal estimation is generally employed to obtain a real-time channel response; although the existing channel estimation method has higher estimation precision, the complexity is higher, the estimation time is longer, and the realization is difficult; therefore, in order to satisfy the requirement of obtaining channel estimation values in a short time under the multipath fading channel, there is a need for an improved channel estimation method for an orthogonal frequency division multiplexing communication system.

In OFDM systems, pilot-based channel estimation and compensation is done in the frequency domain. Because the channel response values at the data carriers are achieved by interpolating the channel response values at the pilot carriers, the system performance depends largely on the accuracy of the pilot carrier channel response estimates. The channel estimation based on the pilot frequency has the methods of simpler LS estimation, more complex MMSE estimation, time domain and frequency domain two-dimensional joint estimation and the like. LS estimation is simple to implement, but the estimation precision is not high; MMSE estimation and other complex estimation methods can achieve higher estimation accuracy, but the algorithm is complex and is not easy to implement. Therefore, in order to achieve effective combination of performance and implementation method, different simplification and approximation algorithms of MMSE are proposed in various documents. The LMMSE estimation and filter interpolation channel estimation method can achieve certain estimation accuracy and is convenient to implement.

Disclosure of Invention

The invention aims to overcome the contradiction between the estimation precision and the realization complexity of the existing channel estimation method of the OFDM communication system, and provides a channel estimation method in the OFDM communication system, which can meet certain estimation precision and is convenient to realize.

The channel estimation method in the orthogonal frequency division multiplexing communication system comprises the following steps:

(1) generating an orthogonal frequency division multiplexing pilot signal for channel estimation by a system transmitter according to a pilot insertion mode and the selected pilot value; the generated pilot signal is converted into a time domain pilot signal through inverse Fourier transform;

(2) transmitting the time domain pilot signal on a wireless multi-path fading channel;

(3) the system receiver performs Fourier transform on the time domain pilot signal to obtain a frequency domain pilot signal; performing channel estimation on a wireless channel, firstly performing least square estimation on a frequency domain pilot signal to obtain a rough estimation value of a sub-channel where a pilot frequency is located, and then estimating the rough estimation value again by using a linear minimum mean square error estimation method to obtain a fine estimation value of the sub-channel where the pilot frequency is located; then, zero value interpolation and low-pass filtering are sequentially adopted to obtain an estimated value of the whole wireless channel.

The method can effectively estimate the channel response in the OFDM communication systemThe method provides a feasible method for solving the contradiction between the improvement of the estimation precision and the reduction of the calculation complexity of the existing channel estimation. FIG. 4 shows a comparison graph of the minimum mean square error of the LMMSE estimation MSE under different delay Rayleigh multipath channels, and it can be seen from the graph that the minimum mean square error performance index of the LMMSE estimation is 10^-2And 10^-3And the effect is ideal, and the estimation precision requirement can be met. Fig. 5 shows the comparison of the error rate performance between the channel estimation method of LMMSE estimation filtering interpolation and the channel estimation method of LS estimation linear interpolation under the Rayleigh multipath channel. As can be known from fig. 5, the LMMSE estimation filtering interpolation method has a signal-to-noise ratio improvement of 0.2 to 4dB, which is more than 1dB in most cases, than the LS estimation linear interpolation method with the same bit error rate, and the relative performance is more ideal. The method has the advantages that theoretically, for pilot signals passing through a multipath fading channel, the LS estimation method is greatly influenced by inter-channel interference ICI and white Gaussian noise AWGN, so that the channel estimation precision is not high; the LMMSE estimation utilizes the autocorrelation statistical property of the channel and the prior knowledge of the SNR (signal to noise ratio) on the basis of the LS estimation, and can overcome the influence of inter-channel interference ICI and Gaussian white noise AWGN to a certain extent, so that the accuracy of the channel estimation of the pilot frequency is greatly improved. In addition, the filter interpolation has better effect than simple linear interpolation filtering and smoothing, so the effect of the whole channel estimation can be obviously improved from 0.2 to 4 dB.

Drawings

Fig. 1 is a block diagram of an implementation of the channel estimation method of the present invention;

FIG. 2 is a block diagram of an apparatus for implementing LMMSE estimation;

FIG. 3 is a block diagram of an apparatus for implementing filter interpolation;

FIG. 4 is a diagram of a comparison of LMMSE estimated MSE minimum mean square error under different delay Rayleigh multipath channels;

fig. 5 is a comparison of the error rate performance between the channel estimation method of LMMSE estimation filtering interpolation and the channel estimation method of LS estimation linear interpolation under the Rayleigh multipath channel.

Detailed Description

Firstly, an OFDM pilot signal used for channel estimation is generated by a transmitter, a comb pilot insertion mode is adopted as a pilot insertion mode, and a point on a QPSK constellation diagram is selected as a constant pilot by a pilot value. The generated pilot signal is transformed into a time domain pilot signal through inverse fourier transform (IFFT).

The transmitter transmits the time domain pilot signal onto a wireless multi-path fading channel.

In the receiver, the channel estimation method and implementation are as follows:

the received time domain pilot signal is changed into a frequency domain pilot signal after Fourier transform (FFT), firstly, the frequency domain pilot signal is estimated by an LS least square method to obtain a rough estimation value of a sub-channel where the pilot is located, and then the rough estimation value is estimated again by an LMMSE linear minimum mean square error estimation method to obtain a fine estimation value of the sub-channel where the pilot is located; and obtaining the frequency response of the whole channel by adopting a filter interpolation method for the fine estimation value, and finishing the whole process of channel estimation.

The principle of the method of the invention is as follows:

the time domain signal y (n) of the OFDM received by the receiver is generated by the combined action of the transmitted signal, the channel transfer function and the white gaussian noise:

y(n)＝x(n)*h(n)+w(n)

the representation in the frequency domain after fourier transformation is,

Y(k)＝X(k)H(k)+I(k)+W(k)

the purpose of channel estimation is to estimate h (k) from the received signal.

For pilot signals that are

Y_P(k)＝X_P(k)H_P(k)+I_P(k)+W_P(k)

X_p(k) For known pilot signal values, Y_p(k) For the received signal value, H_p(k) Is a value of the channel frequency response, W, on a subcarrier_p(k) Is white Gaussian noise AWGN, I on subcarrier_p(k) Is the inter-sub-channel interference ICI.

Pilot-based channel estimation can be divided into two steps, first estimating H_p(k) Then to H_p(k) Interpolation is carried out to recover the value of H (k). The implementation block diagram of the present invention is shown in fig. 1, firstly, LMMSE estimation is performed on the sub-channel where the pilot is located at 101, and then, at 102, the estimated value of the whole channel is obtained by interpolation of a filter.

The specific implementation steps and methods are described as follows:

after the time domain signal is converted into a frequency domain signal by fourier transform,

(1) firstly, LS (least squares) least square method is used for obtaining a rough estimation value of channel impulse response at a receiver of an OFDM communication system

{\hat{H}}_{p, ls} = {[H_{p, ls} (0), H_{p, ls} (1), . . ., H_{p, ls} (N_{p} - 1)]}^{T}

= X_{P}^{- 1} Y_{P}

= [\frac{Y_{P} (0)}{X_{P} (0)}, \frac{Y_{P} (1)}{X_{P} (1)}, . . ., \frac{Y_{P} (N_{P} - 1)}{X_{P} (N_{P} - 1)}]

Wherein X_p(0)，X_p(1)，…，X_p(N_p-1) is a transmitted pilot, N_pIs the number of pilots transmitted. i.e. i₀，…，i_N0-1Is the corresponding position of the pilot signal on the subcarrier.

(2) Then, based on LS coarse estimation, LMMSE (Linear MMSE) estimation is used for obtaining a fine estimation value of channel impulse response

Using static pilot values X by averaging the transmitted signals_P(m)＝c，m＝0，1，...，N _P1 to obtain a simplified Linear MMSE estimation method,

{\hat{H}}_{P} = R_{H_{p} H_{p}} {(R_{H_{p} H_{p}} + \frac{β}{SNR} I)}^{- 1} {\hat{H}}_{p, ls}

wherein,

R_{H_{p} H_{P}} = E {H_{p} H_{p}^{H}},

the autocorrelation matrix is a channel response autocorrelation matrix, and can be obtained by the actual measurement of a receiver or the prior statistical measurement by transmitting a training sequence by a transmitter;

SNR = E | X_{P} (k) |^{2} / σ_{n}^{2}

the signal-to-noise ratio of the pilot carrier signal can be obtained by real-time measurement at a receiver; β ═ E | X_P(k)|²·E|1/X_P(k)|²For QPSK modulation, β is 1, a constant that depends on the modulation signal constellation. The specific implementation and the device block diagram are shown in fig. 2.

(3) Finally, the fine estimation value of the channel impulse response obtained by LMMSE is subjected to filter interpolation (Filter interpolation) to obtain the impulse response value of the whole channel

The implementation of the filter interpolation and the device block diagram are shown in fig. 3, firstly, the interpolation with integral multiple of I is performed on two adjacent pilot frequency estimated values, i.e. I-1 0 value points are inserted between two adjacent pilot frequencies, and then low-pass filtering is performed, so that the channel response value of the data carrier between two adjacent pilot frequencies is obtained through interpolation, and the impulse response value of the whole channel is obtained.

Theoretically, analytically, MMSE estimation method

{\hat{H}}_{p, MMSE} = R_{H_{p} H_{p, ls}} R_{H_{P} H_{P, ls}}^{- 1} {\hat{H}}_{p, ls}

= R_{H_{p} H_{p}} {(R_{H_{p} H_{p}} + σ_{n}^{2} {(X_{p} X_{p}^{H})}^{- 1})}^{- 1} {\hat{H}}_{p, ls}

Matrix inversion operation and matrix multiplication operation are required to be carried out twice in total, and the calculated amount and the processing are complex; the LMMSE estimation method only needs one time of matrix inversion operation and matrix multiplication operation, and the processing complexity is greatly reduced on the basis of obtaining the estimation effect which is closer to the MMSE.

The present invention will be described in further detail below with reference to examples.

Firstly, the transmitter generates an OFDM pilot signal used for channel estimation, the pilot insertion mode adopts a comb pilot insertion mode in which the pilot and data are 1: 3, the pilot value selects a point on a QPSK constellation, and in this embodiment, 1+ j is taken as the pilot signal. The generated pilot signal is transformed into a time domain pilot signal through inverse fourier transform (IFFT).

In the receiver, the channel estimation method and implementation are shown in fig. 1:

the receiver performs Fourier transform (FFT) on the time domain pilot signal to change the time domain pilot signal into a frequency domain pilot signal, and extracts the pilot signal from a pilot insertion position which is known in advance; the pilot signal is input to a 101LMMSE estimation module to obtain a channel fine estimation value, the channel fine estimation value is input to a 102 filter for interpolation, and an impulse response value of the whole channel is output.

The implementation process of the 101LMMSE estimation module corresponding to fig. 1 is shown in fig. 2, wherein a statistical prior channel impulse response value 201 is divided into two paths, one path is transposed in a 202 matrix arithmetic unit and is multiplied by the other unprocessed path in a 203 multiplier after conjugation is solved to obtain a channel autocorrelation value, and a channel autocorrelation matrix R is obtained after averaging is solved by a 204 weighted accumulator_HpHpThe identity matrix stored with 205 is multiplied by the gain measured by the receiver at 206The products of (D) are added in an adder 207, and the added sum is subjected to a matrix arithmetic unit 208 to obtain an inverse matrix; 209 the received pilot signal is divided by reference pilot value 210 (1 + j in this example) at 211 to obtain the coarse estimate of the LS channel; 204 of the channel autocorrelation matrix R_HpHpThe result obtained by the step 208 and the LS channel coarse estimation value of the step 211 are multiplied by a multiplier 212 to obtain an LMMSE channel fine estimation value, and the LMMSE channel fine estimation value is output by a multiplier 213.

Corresponding to the interpolation method of the filter 102 in fig. 1, as shown in fig. 3, for the pilot subcarrier channel estimation values output by 101 in fig. 1, first, 4 times of upsampling is performed in the zero value interpolator 301 in fig. 3, that is, three zeros are inserted between every two pilot subcarrier channel estimation values, and the upsampled signal sequence is filtered and interpolated by a low pass filter (i.e., LPF)302 to obtain an impulse response value of the entire channel, where the LPF is a 51-order FIR low pass filter in this embodiment.

In the method of the invention, the pilot frequency inserting mode can be a plum blossom-shaped pilot frequency inserting mode in which the pilot frequency positions of front and back time slots are staggered, besides the comb-shaped pilot frequency with fixed position. Simulation test results show that the ratio of pilot signal to data signal (1: n) can be 1: 3, or 1: 4 or 1: 5 under the condition of high signal-to-noise ratio (for example, more than 8dB) according to the difference of signal-to-noise ratio of wireless multi-path channel.

The LMMSE estimation, which is the main part of the channel estimation method herein, is a simplified approximation method of MMSE estimation; the invention can obtain the effect of approximate channel estimation for other linear (such as polynomial approximation algorithm of MMSE) or nonlinear MMSE estimation methods.

The interpolation filter used by the invention is an FIR filter with higher order and better linear phase; when the channel condition is good, the receiver can easily compensate the phase offset of the channel estimation or the implementation complexity is low, an IIR filter or other filter with some linear phase difference but a low order may be used.

Claims

1. The channel estimation method in the orthogonal frequency division multiplexing communication system comprises the following steps:

2. The method of claim 1, wherein: and (3) adopting a comb pilot frequency insertion mode as the pilot frequency insertion mode in the step (1).

3. The method of claim 2, wherein: in the comb-shaped pilot insertion method, the ratio of the pilot to the data signal is 1 to n, and n is 3, 4, and 5.

4. A method according to claim 1, 2 or 3, characterized in that: the low-pass filtering in the step (3) adopts a finite impulse response filter or an infinite impulse response filter.

5. A method according to claim 1, 2 or 3, characterized in that: the zero value interpolation is to insert n-1 zeros between two adjacent pilot subcarrier channel estimates, where n is a multiple of the data signal over the pilot signal.