CN101888352A - Channel Estimation and Equalization Method for Suppressing Long Echo and High Doppler in DTMB System - Google Patents

Abstract

The invention belongs to the technical field of wireless digital communication and broadcasting, in particular to a channel estimation and equalization method for suppressing long echo and high Doppler in a DTMB system. On the basis of obtaining the channel impulse response of the current frame through rough estimation, the present invention eliminates the smearing introduced by the frame header and frame body after passing through the channel by reconstructing and equalizing the previous frame body, so as to accurately estimate the channel impulse response Estimation, the obtained channel impulse response can not only reflect the rapid change of the channel, but also eliminate the error caused by the long echo. Finally, the frame body impulse response obtained by simple average is used to reconstruct and equalize the previous frame, and get Output data, thereby eliminating the impact of long multipath and high-speed mobile environments on transmitted data. Computer simulations show that the method has excellent channel estimation performance and moderate complexity, supports reception in high-speed mobile environment and can resist ultra-long echo.

Description

Channel Estimation and Equalization Method for Suppressing Long Echo and High Doppler in DTMB System

technical field

The invention belongs to the technical field of wireless digital communication and broadcasting, and in particular relates to a channel estimation and equalization method oriented to the national standard for digital television terrestrial transmission.

Background technique

The time-domain synchronous OFDM technology proposed by Tsinghua University is an important part of China's digital TV terrestrial transmission standard (DTMB). The standard was proposed in 2006, including two modes of single carrier and multi-carrier. Among them, the multi-carrier mode uses the orthogonal frequency division multiplexing technology to modulate the frame body. Compared with the traditional frequency division multiplexing, this technology improves the spectrum utilization rate and has excellent anti-multipath ability. This mode uses the PN sequence and its cyclic prefix and cyclic suffix as the frame header, which not only acts as a guard interval, but also is used for synchronization and channel estimation. This approach makes it unnecessary to insert pilots in the frequency domain, thereby also improving spectrum utilization (refer to the Chinese invention patent applied by Tsinghua University, application number CN01124144.6).

DTMB standard constellation mapping method can choose 4QAM, 4QAM-NR, 16QAM, 32QAM and 64QAM, 4QAM is suitable for high-speed mobile occasions, while 32QAM and 64QAM support high data rate services. The frame header mode of the DTMB system is divided into PN420 mode in multi-carrier mode, PN945 mode and PN595 mode in single-carrier mode. The PN420 mode consists of a PN255 sequence and its pre-synchronization and post-synchronization. The PN255 sequence length is 255 and is generated by an 8-order LFSR. However, in order to reduce the correlation of adjacent frame headers, the order of the PN sequence is also specified. Reset at the beginning of every superframe. The average power of the frame header data is twice that of the frame body, using BPSK modulation. The pre-synchronization is composed of the last 82 data of the PN sequence, and the post-synchronization is the first 83 data information of the PN sequence. The frame header formed in this way has quasi-cyclic characteristics, and after design, the number of 1s in the frame header is more than 0 and only 1 more. The PN945 mode is roughly the same as the PN420 mode, except that the PN511 sequence and its front and back synchronization are used. The PN511 sequence is generated using a 9th-order LFSR. The PN595 mode does not use the cyclic PN sequence, but intercepts the first 595 bits of the 10-order binary pseudo-random sequence, resets at the beginning of the signal frame, and uses the same sequence for each superframe. In addition to the aforementioned signal frame and super frame, there are sub-frames and daily frames. Since the transmission system is synchronized with the normal time, the daily frame corresponds to 24 hours, the divided frame corresponds to 1 minute, the super frame corresponds to 125 milliseconds, and the signal frame corresponds to different time lengths according to the length of the frame header. In the frame body data processing step, the multi-carrier mode can be directly output, and for the single-carrier mode, it needs to perform inverse discrete Fourier transform after interleaving in the frequency domain and convert it into a time domain signal. (Refer to Chinese National Standard GB20600-2006, Digital TV Terrestrial Broadcasting Transmission System Frame Structure, Channel Coding and Modulation, China Standard Press, 2006)

In wireless communication systems, due to the influence of channel fading, the received data has experienced fading effects such as multipath delay and Doppler frequency shift, so the obtained data has been severely damaged. At this time, if such data is directly used for decoding, valid data cannot be obtained. In order to solve this problem, we must try to eliminate this influence of the channel. A relatively simple method is to include some known information in the received data, and compare this part of the known information with their values after passing through the channel to obtain channel information, and then eliminate the influence of the channel in the transmitted data.

This part of known information may be a pilot in the frequency domain or a training sequence in the time domain. For multi-carrier systems, frequency domain division can be used to achieve channel equalization, and for single-carrier systems, some adaptive equalization techniques are generally used. For the DTMB system, the known information uses the time-domain training sequence on the frame header, and we use this part of the information to obtain the channel impulse response, and then perform channel equalization. Channel equalization is implemented using fast Fourier transform. Compared with multi-carrier mode, if frequency domain equalization is used in single carrier mode, an IFFT needs to be performed to convert back to the time domain.

In the DTMB system, the simplest channel estimation method is to use the quasi-cyclic characteristics of the frame header PN sequence in the multi-carrier mode, and only need a simple time-domain correlation to obtain the channel impulse response (refer to the Chinese invention patent "time-based Domain-related terrestrial digital multimedia broadcasting system channel estimator", authorization number is 200710044718.0), and this method also has better mobile reception performance, but when the channel length is large, this method shows its disadvantages, In addition, this algorithm is not suitable for single-carrier mode and does not meet the design of the national standard fusion standard. At this time, we can use the iterative elimination algorithm to perform channel estimation (refer to the Chinese invention patent "An Iterative Elimination Method for Pseudo-random Sequence Filling in OFDM Modulation System" applied by Tsinghua University, the authorization number is 200510012127.6, and the document F.Yang etal.Channel Estimation for the Chinese DTTB System Based on a Novel Iterative PNSequence Reconstruction, ICC, 2008, 54(4): 1583-1589), this method has a great advantage in combating long echoes, but these algorithms cannot It supports high-speed mobile reception. In addition, there is an algorithm that has a certain performance in supporting long echoes, but cannot support ultra-long echoes. In the case of short channel lengths, this algorithm degenerates into a simple PN sequence correlation algorithm (reference L. Gui, Q. Li, B. Liu et al. Low Complexity Channel Estimation Method for TDS-OFDM Based Chinese DTTB System. IEEE TRANSACTIONS ONCONSUMER ELECTRONICS, 2009, 55(3): 1135-1140).

In the background of the above technologies, in order to satisfy the high-speed mobile reception of the DTMB system and combat the ultra-long echo at the same time, we propose a new channel estimation and equalization technology.

Contents of the invention

The purpose of the present invention is to provide a DTMB system-oriented method for channel estimation and equalization that can simultaneously satisfy long-echo and high-Doppler applications.

Denote the unmodulated frame body data of the DTMB system baseband as S _i , where i represents the frame number.

For OFDM systems, this data needs to be modulated, so

${\mathrm{the\; s}}_{i,k}=\frac{1}{\sqrt{{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="HSA00000182111300012.png,HSA00000182111300021.png"\; state="\{\{state\}\}">N</figure-callout>}}_2}}\underset{\mathrm{no}=0}{\overset{N_2-1}{\mathrm{\&Sigma;}}}{S}_{i,\mathrm{no}}\exp \left(\frac{2\mathrm{\&pi;kn}}{N_2}\right),$ 0≤k＜3799

Here, it means that the data is modulated by IDFT, N ₂ is the number of subcarriers 3780, and the range of k also indicates the length of the signal.

In addition, a frame header needs to be added. For the multi-carrier mode, the frame header is the PN sequence plus pre-sync and post-sync, but for the single-carrier mode, it is the first 595 data values of the m-sequence. Here, the frame header data is denoted as c _i , without specifying which mode it corresponds to. Variables in uppercase are used to represent frequency-domain values, while variables in lowercase are used to represent time-domain values.

The channel is processed using the tapped delay line model, where each tap contains a colored Gaussian noise that conforms to the Jakes or Gauss Doppler spectrum. Here, an appropriate number of taps is selected so that the channel can meet the actual situation.

After the transmitted signal passes through the channel, due to multipath delay, the transmitted data does not arrive at the same time, and the data that arrives later will pollute the subsequent data. We obtain the position of the main path through synchronization, and record the frame header corresponding to the main path as y _i , where i also represents the frame number, which is calculated by synchronization, and _xi represents the frame body part corresponding to the received main path part. Here, y _i contains the tail from the i-1 frame header, and _xi Then contains the tailing from the frame header of the i-th frame. The data frame structure at the transmitting end is shown in Figure 1(a), and the received signal frame structure after passing through the channel is shown in Figure 1(b).

In addition, since we assume that the channel is fast time-varying, the difference between the channel impulse response of each frame cannot be ignored, even the difference between the header and the frame body of the same frame cannot be ignored. Then x _i can be expressed as:

${\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}{<span\; class="notranslate">\; \{</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; \&CircleTimes;</span>\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; \}</span>}_{\mathrm{<span\; class="notranslate">\; 0</span>}}^{\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; +</span>{<span\; class="notranslate">\; \{</span>{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; \&CircleTimes;</span>\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; \}</span>}_{\mathrm{<span\; class="notranslate">\; <figure-callout\; id="2"\; label="N\; N"\; filenames="HSA00000182111300012.png,HSA00000182111300021.png"\; state="\{\{state\}\}">N</figure-callout></span><figure-callout\; id="2"\; label="N\; N"\; filenames="HSA00000182111300012.png,HSA00000182111300021.png"\; state="\{\{state\}\}">\; </figure-callout>}}^{{\mathrm{<span\; class="notranslate">\; </span>}}_{}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; w</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; L</span>}\\ {<span\; class="notranslate">\; \{</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; \&CircleTimes;</span>\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; \}</span>}_{\mathrm{<span\; class="notranslate">\; L</span>}}^{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; w</span>}& \mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; m</span>}\end{array}\right.$

above, Represents convolution operation, w represents additive Gaussian white noise, Indicates the first L data of the convolution result of the data sent by the header of the current frame and the channel impulse response. Similar are:

${\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}{<span\; class="notranslate">\; \{</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; \&CircleTimes;</span>\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; \}</span>}_{\mathrm{<span\; class="notranslate">\; 0</span>}}^{\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; +</span>{<span\; class="notranslate">\; \{</span>{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; \&CircleTimes;</span>\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; \}</span>}_{\mathrm{<span\; class="notranslate">\; N</span>}}^{\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; w</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; L</span>}\\ {<span\; class="notranslate">\; \{</span>{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; \&CircleTimes;</span>\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; \}</span>}_{\mathrm{<span\; class="notranslate">\; <figure-callout\; id="2"\; label="L\; N"\; filenames="HSA00000182111300012.png,HSA00000182111300021.png"\; state="\{\{state\}\}">L</figure-callout></span><figure-callout\; id="2"\; label="L\; N"\; filenames="HSA00000182111300012.png,HSA00000182111300021.png"\; state="\{\{state\}\}">\; </figure-callout>}}^{{\mathrm{<span\; class="notranslate">\; </span>}}_{}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; w</span>}& \mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; <figure-callout\; id="2"\; label="N"\; filenames="HSA00000182111300012.png,HSA00000182111300021.png"\; state="\{\{state\}\}">N</figure-callout></span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\end{array}\right.$

In the present invention, three frame header modes can be supported by changing the frame header data, and the processing flow of single carrier is basically the same as that of multi-carrier. i represents the current frame, i-1 represents the previous frame, and so on. M represents the length of the frame header, N ₂ represents the length of the frame body, and N ₂ can be 3780. N ₁ is one of the FFT lengths, N ₁ can be 2048, and L represents the length of the tail. x _i and y _i represent the frame body and frame header of the received data, respectively.

The present invention can use digital signal processor, in the realization of programmable logic device or application-specific integrated circuit, concrete steps are as follows:

(1) Step 1: Roughly estimate the frame header of the i-th frame

First, the channel impulse response (CIR) of the current frame (frame i) is calculated. The method adopted is to perform N ₂ -point FFT on the N ₂ data (r _headi ) including the frame header, and at the same time perform N ₂ -point FFT on the locally stored frame header information ( _ci ), and divide the two, By performing IFFT again, the time-domain impulse response of the channel in frame i-1 can be obtained. The value of the impulse response at this time is not an accurate value because it includes the tailing of the frame header and the tailing of the frame body, but we will try to eliminate the influence of these tails later, so as to obtain an accurate impulse response. (see steps 2, 3, 4). See Figure 2(a).

First we write the definition of r _headi

${\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; head</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}& \mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; m</span>}\\ {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; m</span>}}& \mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; L</span>}\end{array}\right.$

Here L has been known before, if it is the first frame, then take the maximum possible value of the channel, namely:

$\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>& \mathrm{<span\; class="notranslate">\; forfram</span>}{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}\\ {\mathrm{<span\; class="notranslate">\; L</span>}}_{\mathrm{<span\; class="notranslate">\; estimated</span>}}<span\; class="notranslate">\; ,</span>& \mathrm{<span\; class="notranslate">\; for\; other\; cases</span>}\end{array}\right.$

Then the roughly estimated process is

$h_{i,k}=\mathrm{IFFT}\left\{\frac{\mathrm{FFT}\left(r_{\mathrm{head},k}\right)}{\mathrm{FFT}\left(c_{i,k}\right)}\right\},$ 0≤k<N ₁

After the rough estimation is completed, a dynamic threshold filtering and channel length estimation need to be performed. The former is beneficial to suppress the noise generated during the estimation process, and the latter is beneficial to reduce the influence of noise by accurately estimating the channel length. These two parts are discussed after step 4.

(2) Step 2: Equalize i-1 frame and calculate tailing

After the rough estimation of the channel impulse response of the current frame is completed, an equalization needs to be performed on the previous frame (i-1 frame). However, the result obtained by this equalization is not used for output, but to eliminate the effect of tailing, so as to obtain a more accurate impact response. See Figure 2(b).

First, we have obtained the shock response of frame i-1, because the shock response of the current frame has been obtained in the last process, and the response of the "current frame" becomes the CIR of the previous frame in the current process. At present, since the tailing of the header of the i-1 frame in the body of the i-1 frame needs to be eliminated, and the tailing of the frame body of the i-1 frame in the header of the i-th frame also needs to be eliminated, so here we need to use CIR of frame i and frame i-1. This process can be expressed as

u _{headi-1, k} = IFFT {FFT(h _{i-1, k} )×FFT(ci _{-1, k} )}, 0≤k<N ₁

u _{headi-1, k} represents the situation after the local i-1 frame header passes through the channel, and is obtained through N _1- point FFT and IFFT. Through u _headi-1,k , the trailing of the i-1 frame header in the i-1 frame body can be calculated. The subtraction here is only performed on the data within the channel length:

$z_{\mathrm{head}-1,k}=x_{i-1,k}-{\left\{u_{\mathrm{head}-1,k-m}\right\}}_0^{L-1},$ 0≤k<L

Similarly, the situation after the local i-frame header passes through the channel can be calculated:

u _{headi, k} = IFFT {FFT(h _{i, k} )×FFT(ci _{, k} )}, 0≤k<N ₁

And the tailing of the i-1 frame body in the i-th frame header:

$z_{\mathrm{taili},k}={\mathrm{the\; y}}_{i,k}-{\left\{u_{\mathrm{head},k}\right\}}_0^{L-1},$ 0≤k<L

Therefore, this value can be added at the beginning of the frame body to complete the reconstruction of the i-1 frame body:

${\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; bodyi</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}{\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; head</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; taili</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}& \mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}\\ {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}& \mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; <figure-callout\; id="2"\; label="N"\; filenames="HSA00000182111300012.png,HSA00000182111300021.png"\; state="\{\{state\}\}">N</figure-callout></span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\end{array}\right.$

Next, the channel impulse response of the i-1 frame body can be calculated by simple averaging [28]. i.e. using:

${\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; bodyi</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; head</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; <figure-callout\; id="2"\; label="h\; head"\; filenames="HSA00000182111300012.png,HSA00000182111300021.png"\; state="\{\{state\}\}">h</figure-callout></span><figure-callout\; id="2"\; label="h\; head"\; filenames="HSA00000182111300012.png,HSA00000182111300021.png"\; state="\{\{state\}\}">\; </figure-callout>}}_{\mathrm{<span\; class="notranslate">\; </span>}}}{\mathrm{<span\; class="notranslate">\; 2</span>}}$

(Due to the rapid change of the channel, if the CIR estimated by the frame header is used as the CIR of the frame body, the result will be inaccurate)

Then use the CIR obtained above to equalize and convert back to the time domain:

${\widetilde{\mathrm{the\; s}}}_{i-1}=\mathrm{IFFT}\left\{\frac{\mathrm{FFT}\left(r_{\mathrm{bodyi}-1,k}\right)}{\mathrm{FFT}\left(h_{\mathrm{bodyi}-1,k}\right)}\right\},$ 0≤k<N ₂

The above equalization step needs to use two 3780-point FFTs to process the frame body data and the shock response of the frame body respectively, and obtain the equalized data after division. However, this frequency domain data is not needed at this time, so it is still converted to the time domain.

(3) Step 3: Update the channel impulse response of frame i-1

Now, the CIR of frame i-1 can be updated with the previously equalized data. Since the tailing of the i-1 frame body in the i-frame header has been eliminated in step 2 of the previous process, in the current process, it is only necessary to remove the first L and The data value, splicing the M data of the frame header received on the splicing, has just completed the reconstruction of the frame header (see Figure 2 (c)):

$u_{\mathrm{body\; top}-1,k}=\mathrm{IFFT}\{\mathrm{FFT}\left({\widetilde{\mathrm{the\; s}}}_{i-1,k}\right)\&times;\mathrm{FFT}\left(h_{i-1,k}\right)\},$ 0≤k<N ₁

The situation after the i-1 frame body passes through the channel is calculated above, and the u _{body topi-1,k} is subtracted from the received data to obtain the drag of the i-1 frame header in the i-1 frame body tail:

$z_{\mathrm{head}-1,k}=x_{i-1,k}-{\left\{u_{\mathrm{head}-1,k-m}\right\}}_0^{L-1},$ 0≤k<L

and then

${\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; head</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}& \mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; m</span>}\\ {\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; headtop</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; m</span>}}& \mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; <figure-callout\; id="1"\; label="N"\; filenames="HSA00000182111300011.png,HSA00000182111300012.png"\; state="\{\{state\}\}">N</figure-callout></span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}\end{array}\right.$

Then the updated impulse response can be obtained:

$h_{i-1,k}=\mathrm{IFFT}\left\{\frac{\mathrm{FFT}\left(r_{\mathrm{head}-1,k}\right)}{\mathrm{FFT}\left(c_{i-1,k}\right)}\right\},$ 0≤k<N ₁

Similarly, this CIR is also filtered by a dynamic threshold, but there is no need to perform channel length estimation. Compared with step 1, the result of dynamic threshold filtering at this time is more accurate, so different coefficients can be used to reduce noise filtering and obtain better precision.

(4) Step 4: Equalize the i-2 frame as the final output

After the updated i-1 frame header is obtained, the i-1 frame CIR in the previous process becomes the i-2 frame CIR in the current process, so the final i-2 is obtained through simple averaging Frame body CIR (see Figure 2(d)):

${\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; bodyi</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; head</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; head</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}}{\mathrm{<span\; class="notranslate">\; 2</span>}}$

Therefore, using the updated hi _-1,k , the frame body of frame i-2 can be rebalanced

${\tilde{S}}_{i-2}=\frac{\mathrm{FFT}\left(r_{\mathrm{bodyi}-2,k}\right)}{\mathrm{FFT}\left(h_{\mathrm{bodyi}-2,k}\right)},$ 0≤k<N ₂

是i-2帧帧体经过均衡后的频域。接下来就可以读取下一帧进行处理，只要重复前面的4个步骤就可以了。At this point, a processing process is completed.

is the equalized frequency domain of the i-2 frame body. Then you can read the next frame for processing, just repeat the previous 4 steps.

In the present invention, described dynamic threshold filtering, method is as follows:

Let h be the estimated channel impulse response, so we denote the real part of h as h _real , then the value after filtering is [7]

${\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; real</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; real</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; real</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; SNR</span>}}}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; real</span>}}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; adj</span>}$

Among them, SNR represents the signal-to-noise ratio, and adj is the parameter value. Both values can be calculated by the following formula

$\mathrm{<span\; class="notranslate">\; adj</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; max</span>}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; q</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; SNR</span>}}}{\mathrm{<span\; class="notranslate">\; max</span>}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; q</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; \&alpha;</span>}<span\; class="notranslate">\; \&Element;</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 0.9,0.99,0.998,0.999</span>}<span\; class="notranslate">\; )</span>$

Choose the value of α according to your needs. For rough estimation, choose a smaller α value to filter out more noise, and for an updated CIR, choose a larger α value to retain more details and improve the accuracy of the equalization result.

The channel length estimation method is as follows:

The purpose of channel length estimation is to specify how long the tail is for subsequent processing, so as to avoid introducing too much noise in the processing (the part exceeding the channel length can be set to 0), the criterion for channel length estimation is to use The main path shall prevail, and the position of the farthest path with power within -20dB of the main path plus 10 shall be used as the channel length, which can not only reduce noise, but also prevent the change of path length from being missed. It is assumed here that the change speed of the delay of the diameter is much smaller than the change speed of the amplitude and phase of the diameter.

In addition, for the initial several frames, all the above-mentioned steps cannot be performed, and appropriate arrangements need to be made to gradually include the above-mentioned steps. For example, for the first frame received, only the first step needs to be performed. In the second frame, the second step can be included, in the third frame, the third step is included, all steps are included from the fifth frame, and the equalized data is output from the sixth frame.

Next, we perform a floating-point simulation of the algorithm. The main simulation parameters are shown in Table 1. In addition, ideal synchronization is assumed here. The channel models we choose include the CT8 channel model, which contains a 0dB echo up to 31.8us (see Table 5, refer to J.Wang et al.Iterative padding subtraction of the PN sequence for theTDS-OFDM over broadcast channel s .IEEE TRANSACTIONS ON CONSUMER ELECTRONICS, 2005, 51(4): 1148-1152). In addition, there are TU6 channel model (see Table 2), DVB P1 channel model (see Table 4), Brazil E channel model (see Table 3) and SFN channel model (see Table 6, refer to B.W.Song et al.On channel estimation and equalization in TDS-OFDM based terrestrial HDTV broadcasting system. IEEE TRANSACTIONS ON CONSUMER ELECTRONICS, 2005, 51(3): 790-797). Since the frame header mode of PN420 has the weakest ability to resist long echo, here we use this mode to test the anti-long echo ability of the proposed algorithm. We used two maximum Doppler shifts of 10Hz and 200Hz. The normalized Doppler frequencies fdTs were 0.05 and 1, respectively. All simulations assume perfect synchronization.

First, we choose 10Hz Doppler frequency shift, and reference algorithm (see literature F. Yang et al. For comparison, QPSK and 16QAM are selected for the constellation mapping mode, and the signal-to-noise ratio ranges from 5dB to 30dB, and the unsigned rates of the two algorithms are compared. As shown in Figure 4, we observe the situation under 200Hz Doppler. Constellation mapping modes and SNR ranges are the same as before. It can be found that the proposed algorithm has great advantages, and can be received normally under both QPSK and 16QAM, while the performance of the previous reference algorithm deteriorates greatly and cannot work normally. Compared with low Doppler, the performance of the proposed algorithm is lost, but it is still within an acceptable range. In addition, the main reason for performance degradation is that simple interpolation under high-speed movement can no longer reflect the actual response of the channel well.

Next, as shown in Figure 5, in the case of low Doppler, we compare the performance of the three modulation methods of QPSK, 16QAM and 64QAM under the CT8 channel with the reference method. It can be found that the unsigned rate performance under QPSK is the best, while the performance under 64QAM is relatively poor. This is because the modulated data is more sensitive to noise and channel estimation errors in 64QAM mode, but from the simulation results, it is still is in line with the requirements.

Finally, we compare the performance of the proposed algorithm under other channel conditions, as shown in Fig. 6. Including TU6 channel, DVB P1 channel and Brazil-E channel, and also simulated the performance of long echo under 50us (SFN channel). The Doppler frequency shift is 10Hz, and the modulation method is 16QAM. Among these channels, the worst performance is DVB-P1 channel, followed by SFN channel and CT8 channel. DVB P1 channel performance loss is due to the large number of multipaths, and SFN channel is due to the existence of ultra-long multipaths.

Table 1 DTMB system simulation parameters

project parameters symbol rate 7.56×10 ⁶ Constellation Mapping Mode QPSK, 16QAM, 64QAM

project parameters FFT length N ₁ 1024 FFT length N ₂ 3780 Frame header mode 420 carrier mode Multi-carrier

Table 2 TU6alt channel model parameters

tap Delay (us) Gain(dB) 1 0 -3 2 0.2 0 3 0.5 -2 4 1.6 -6 5 2.3 -8 6 5.0 -10

Table 3 Brazil E-channel model parameters

Tap Delay (us) Gain(dB) 1 0 0 2 1 0 3 2 0

Table 4 DVB P1 channel model parameters

tap Delay (us) Gain(dB) 1 1.003 -24.7822 2 5.4221 -15.0499 3 0.5187 -7.8046 4 2.7518 -10.3544 5 0.6029 -11.7413 6 1.0166 -24.1759 7 0.1436 -16.4585 8 0.1538 -25.7581 9 3.3249 -14.6531 10 1.9356 -7.9378 11 0.4299 -10.5823 12 3.2289 -9.0982 13 0.8488 -11.6039 14 0.0739 -12.9219 15 0.204 -15.3403 16 0.1942 -16.4942 17 0.9244 -12.3907 18 1.3813 -18.5193 19 0.6405 -13.1061 20 1.3687 -11.7096

Table 5 CT8 channel model parameters

Tap Delay (us) Gain(dB) 1 0 0 2 -1.8 -18 3 0.15 -20 4 1.8 -20 5 5.7 -10 6 31.8 0

Table 6 SFN channel model parameters

Tap Delay (us) Gain(dB) 1 0 0 2 19 -5 3 52 -1

Description of drawings

FIG. 1 is a schematic diagram of the frame structure of the sending end and the receiving end.

Figure 2 is an illustration of the invention.

Fig. 3 is a block diagram of the processing flow of the present invention.

FIG. 4 is a comparison of bit error rates between the present invention and the reference method under different signal-to-noise ratios and Doppler frequency shifts in the CDT8 channel.

Fig. 5 is a comparison of bit error rates between the present invention and the reference method under different signal-to-noise ratios and modulation modes in CDT8 channel.

FIG. 6 is a comparison of bit error rates of the present invention under various channel conditions.

Detailed ways

The present invention proposes a channel estimation and equalization method oriented to China's terrestrial digital television transmission standard. The channel estimation and equalization steps of the present invention will be described below in conjunction with FIG. 3 . Define a processing flow as the following 1) to 8), each processing flow will output a frame of processed data, and return to 1) after completing a processing flow.

1) During the process of channel estimation, we need to fully cache three frames of data, and also need to store the frame headers of the previous two frames. First we need to calculate the Channel Impulse Response (CIR) of the current frame (frame i). The method adopted is to perform N ₂ -point FFT on the N ₂ data (r _headi ) including the frame header, and at the same time perform N ₂ -point FFT on the locally stored 420-length frame header information (c _i ). The time-domain impulse response of the channel in frame i-1 can be obtained by performing IFFT again.

${\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; head</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}& \mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; m</span>}\\ {\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; m</span>}}& \mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; L</span>}\end{array}\right.$

If it is the first frame, then take the maximum possible value of the channel, that is

$\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; ,</span>& \mathrm{<span\; class="notranslate">\; forfram</span>}{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}\\ {\mathrm{<span\; class="notranslate">\; L</span>}}_{\mathrm{<span\; class="notranslate">\; estimated</span>}}<span\; class="notranslate">\; ,</span>& \mathrm{<span\; class="notranslate">\; for\; other\; cases</span>}\end{array}\right.$

Then the roughly estimated process is

$h_{i,k}=\mathrm{IFFT}\left\{\frac{\mathrm{FFT}\left(r_{\mathrm{head},k}\right)}{\mathrm{FFT}\left(c_{i,k}\right)}\right\},$ 0≤k<N ₁

After the rough estimation is completed, a dynamic threshold filtering and channel length estimation need to be performed, which will be described later. At this time, the obtained impulse response also needs to be stored, and needs to be used in the next processing.

2) Through the previous process, the shock response of the previous frame has been obtained. At present, the tailing of the frame header of frame i-1 in the frame body of frame i-1 needs to be eliminated, and the tailing of the frame body of frame i-1 in the header of the i-th frame also needs to be eliminated. For this reason, the impulse response is calculated Convolution with the PN sequence sent by the i-1 frame:

u _{headi-1, k} = IFFT {FFT(h _{i-1, k} )×FFT(ci _{-1, k} )}, 0≤k<N ₁

u _{headi-1, k} represents the situation after the local i-1 frame header passes through the channel, and is obtained through N _1- point FFT and IFFT.

Similarly, the situation after the local i-frame header passes through the channel can be calculated:

u _{headi, k} = IFFT {FFT(h _{i, k} )×FFT(ci _{, k} )}, 0≤k<N ₁

Thus, the tailing of the i-1 frame body in the frame header of the i-th frame is obtained

$z_{\mathrm{taili},k}={\mathrm{the\; y}}_{i,k}-{\left\{u_{\mathrm{head},k}\right\}}_0^{L-1},$ 0≤k<L

3) Through u _{headi-1, k,} we can calculate the tailing of the frame header of the i-1 frame in the frame body of the i-1 frame. The subtraction here is only performed on the data within the channel length

$z_{\mathrm{head}-1,k}=x_{i-1,k}-{\left\{u_{\mathrm{head}-1,k-m}\right\}}_0^{L-1},$ 0≤k<L

Therefore, this value can be added to the beginning of the frame body of the i-1th frame to complete the reconstruction of the frame body of the i-1 frame

${\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; bodyi</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}{\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; head</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; taili</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}& \mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}\\ {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}& \mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; <figure-callout\; id="2"\; label="N"\; filenames="HSA00000182111300012.png,HSA00000182111300021.png"\; state="\{\{state\}\}">N</figure-callout></span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\end{array}\right.$

4) At this time, the channel impulse response of the i-1 frame body is calculated by simple average.

${\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; bodyi</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; head</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; <figure-callout\; id="2"\; label="h\; head"\; filenames="HSA00000182111300012.png,HSA00000182111300021.png"\; state="\{\{state\}\}">h</figure-callout></span><figure-callout\; id="2"\; label="h\; head"\; filenames="HSA00000182111300012.png,HSA00000182111300021.png"\; state="\{\{state\}\}">\; </figure-callout>}}_{\mathrm{<span\; class="notranslate">\; </span>}}}{\mathrm{<span\; class="notranslate">\; 2</span>}}$

Use the CTR obtained above to equalize the i-1 frame body

${\stackrel{<span\; class="notranslate">\; ~</span>}{\mathrm{<span\; class="notranslate">\; the\; s</span>}}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; IFFT</span>}<span\; class="notranslate">\; \{</span>\frac{\mathrm{<span\; class="notranslate">\; FFT</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; bodyi</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; FFT</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; bodyi</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \}</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; <figure-callout\; id="2"\; label="N"\; filenames="HSA00000182111300012.png,HSA00000182111300021.png"\; state="\{\{state\}\}">N</figure-callout></span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}$

The above equalization step needs to use two 3780-point FFTs to process the frame body data and the shock response of the frame body respectively, and obtain the equalized data after division. However, this frequency domain data is not needed at this time, so it is still converted to the time domain.

5) Update the CIR of the i-1 frame by using the previously equalized data. In the current process, it is only necessary to remove the first L length and data value after the frame body of the i-1 frame passes through the channel, and splice the received M data of the frame header to complete the reconstruction of the frame header .

$u_{\mathrm{body\; top}-1,k}=\mathrm{IFFT}\{\mathrm{FFT}\left({\widetilde{\mathrm{the\; s}}}_{i-1,k}\right)\&times;\mathrm{FFT}\left(h_{i-1,k}\right)\},$ 0≤k<N ₁

The above calculates the situation after the i-1 frame body passes through the channel, subtract u _{body top i-1,k} from the received data, and then you can get the i-1 frame header in the i-1 frame body smear

$z_{\mathrm{head}-1,k}=x_{i-1,k}-{\left\{u_{\mathrm{head}-1,k-m}\right\}}_0^{L-1},$ 0≤k<L

In addition, we also need to calculate the trailing u _{body bottom i-1} left by the i-1 frame body in the i-frame header, and then subtract this value from the i-frame header, so that in the latter processing flow , when reconstructing the frame header of i-1 frame, there is no need to consider the tail left by the frame body of the previous frame, we record it as y _{i updated} and store this value.

6) Reconstruct the i-1 frame header, where y _{i-1, k} uses y _{i updated} :

${\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; head</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}& \mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; m</span>}\\ {\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; headtop</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; m</span>}}& \mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; <figure-callout\; id="1"\; label="N"\; filenames="HSA00000182111300011.png,HSA00000182111300012.png"\; state="\{\{state\}\}">N</figure-callout></span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}\end{array}\right.$

7) Then the updated impulse response can be obtained

$h_{i-1,k}=\mathrm{IFFT}\left\{\frac{\mathrm{FFT}\left(r_{\mathrm{head}-1,k}\right)}{\mathrm{FFT}\left(c_{i-1,k}\right)}\right\},$ 0≤k<N ₁

This CIR is also filtered by a dynamic threshold. In addition, h _{i-1, k} also needs to be stored, denoted as h _{i-1 updated} . In addition, in this step, the influence of the i-1 frame header is eliminated from the i-1 frame, denoted as and store this data.

8) Now, the updated i-1 frame header is obtained, and the i-1 frame CIR in the previous process becomes the i-2 frame CIR in the current process, and the final i-2 is obtained through simple averaging Frame Body CIR. Here, h _{head i-2} is the previously saved g _{i-1 updated} , and h _headi-1 is the previously updated h _{i-1 updated} ,

${\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; bodyi</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; head</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; head</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}}{\mathrm{<span\; class="notranslate">\; 2</span>}}$

于是：Therefore, using the updated h _i-1 , the frame body of frame i-2 can be rebalanced, wherein the frame body saved in the previous processing flow is used

then:

${\tilde{S}}_{i-2}=\frac{\mathrm{FFT}\left(r_{\mathrm{bodyi}-2,k}\right)}{\mathrm{FFT}\left(h_{\mathrm{bodyi}-2,k}\right)},$ 0≤k<N ₂

是i-2帧帧体经过均衡后的频域。接下来就可以读取下一帧进行处理，只要重复前面的步骤就可以了。At this point, a processing process is completed,

is the equalized frequency domain of the i-2 frame body. Then you can read the next frame for processing, just repeat the previous steps.

There are a few other processing steps involved in the flow above:

1) Dynamic threshold filtering

Let h be the estimated channel impulse response, denote the real part of h as h _real , and the value after filtering is:

${\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; real</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; real</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; real</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; SNR</span>}}}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; real</span>}}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; adj</span>}$

Among them, SNR represents the signal-to-noise ratio, and adj is the parameter value. Both values can be calculated by the following formula

$\mathrm{<span\; class="notranslate">\; adj</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; max</span>}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; q</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; SNR</span>}}}{\mathrm{<span\; class="notranslate">\; max</span>}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; q</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; \&alpha;</span>}<span\; class="notranslate">\; \&Element;</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 0.9,0.99,0.998,0.999</span>}<span\; class="notranslate">\; )</span>$

Choose the value of α according to your needs. For the rough estimate, choose a smaller α value to filter out more noise, and for the updated C and R, choose a larger α value to retain more details and improve the accuracy of the equalization result.

2) Channel length estimation

The criterion for channel length estimation is to take the main path as the criterion, and add 10 to the location of the farthest path whose power is within -20dB of the main path as the channel length.

In addition, for the initial several frames, all the above-mentioned steps cannot be performed, and appropriate arrangements need to be made to gradually include the above-mentioned steps. For example, for the first frame received, only the first step needs to be carried out. When the second frame is received, the second step can be included. When the third frame is received, the third step is included. From the fifth The frame starts to include all the steps, and the equalized data is output from the sixth frame.

The implementation process of the present invention has been described in detail above with reference to FIG. 3 , but the present invention is not limited to the above embodiments.

Claims (1)

1.DTMB system suppresses long echo and high Doppler channel estimation and equalization method, it is characterized in that, in using digital signal processor, in the realization of programmable logic device or application-specific integrated circuit, concrete steps are as follows: Step 1: Record the current frame as the i-th frame, and roughly estimate the header of the i-th frame To calculate the channel impulse response CIR of the current frame, the method used is to perform FFT of N ₂ points on the N ₂ data r _headi including the frame header, and also perform N ₂ for the locally stored 420-length frame header information c _i Point FFT; divide the two, and then perform an IFFT to obtain the time-domain impulse response of the channel in the i-1 frame: $h_{i,k}=\mathrm{IFFT}\left\{\frac{\mathrm{FFT}\left(r_{\mathrm{head},k}\right)}{\mathrm{FFT}\left(c_{i,k}\right)}\right\},$ 0≤k<N ₁ Then perform a dynamic threshold filtering and channel length estimation; Step 2: Equalize i-1 frames and calculate smearing Eliminate the tailing of the header of frame i-1 in the frame body of frame i-1 and the tailing of the frame body of frame i-1 in the header of the i-th frame, so as to add this value at the beginning of the frame body to complete i Reconstruction of -1 frame frame body, tailing uses FFT to realize convolution calculation: ${\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; bodyi</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}{\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; head</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; taili</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}& \mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}\\ {\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}& \mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\end{array}\right.$ By averaging, calculate the channel impulse response of the i-1 frame body: ${\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; bodyi</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; head</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; head</span>}}}{\mathrm{<span\; class="notranslate">\; 2</span>}}$ Use the CIR obtained above for equalization: ${\widetilde{\mathrm{the\; s}}}_{i-1}=\mathrm{IFFT}\left\{\frac{\mathrm{FFT}\left(r_{\mathrm{bodyi}-1,k}\right)}{\mathrm{FFT}\left(h_{\mathrm{bodyi}-1,k}\right)}\right\},$ 0≤k<N ₂ in, $z_{\mathrm{head}-1,k}=x_{i-1,k}-{\left\{u_{\mathrm{head}-1,k-m}\right\}}_0^{L-1},$ 0≤k<L $z_{\mathrm{taili},k}={\mathrm{the\; y}}_{i,k}-{\left\{u_{\mathrm{head},k}\right\}}_0^{L-1},$ 0≤k<L u _{head i-1, k} = IFFT{FFT(h _{i-1, k} )×FFT(c _{i-1, k} )}, 0≤k<N ₁ u _{head i, k} = IFFT {FFT(h _{i, k} )×FFT(ci _{, k} )}, 0≤k<N ₁ Step 3: Update the channel impulse response of frame i-1 Remove the first L data values after the frame body of the i-1 frame passes through the channel, and splicing the tailing data in the M frame bodies of the received frame header to complete the reconstruction of the frame header. In the calculation, use The frame body after the previous equalization, ${\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; head</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}& \mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; m</span>}\\ {\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; headtop</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; m</span>}}& \mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&le;</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}\end{array}\right.$ Then update the impulse response. Similarly, this CIR is filtered by dynamic threshold; Step 4: Equalize i-2 frames as final output Through averaging, the final i-2 frame body CIR is obtained, and the updated h _i-1,k is used to rebalance the frame body of the i-2 frame as an output; ${\tilde{S}}_{i-2}=\frac{\mathrm{FFT}\left(r_{\mathrm{bodyi}-2,k}\right)}{\mathrm{FFT}\left(h_{\mathrm{bodyi}-2,k}\right)},$ 0≤k<N ₂ At this point, a processing process is completed,

is the equalized frequency domain of the i-2 frame body; then read the next frame for processing, and repeat the previous 4 steps; Wherein, N ₁ is the FFT length, N ₂ represents the frame body length, L represents the length of the tail, M represents the frame header length, xi _and y _i represent the frame body and the frame header of the received data respectively; The dynamic threshold filtering is specifically: Let h be the estimated channel impulse response, and record the real part of h as h _real , then, the value after filtering: ${\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; real</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; real</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; real</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; SNR</span>}}}<span\; class="notranslate">\; \&times;</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; real</span>}}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; adj</span>}$ Here, SNR represents the signal-to-noise ratio, and adj is the parameter value, which is calculated by the following formula: $\mathrm{<span\; class="notranslate">\; adj</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; max</span>}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; q</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; SNR</span>}}}{\mathrm{<span\; class="notranslate">\; max</span>}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; q</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; \&alpha;</span>}<span\; class="notranslate">\; \&Element;</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 0.9,0.99,0.998,0.999</span>}<span\; class="notranslate">\; )</span>$ For the rough estimate, select the smaller α value, and for the updated CIR, select the larger α value; The channel length estimation is specifically: taking the main path as the criterion, and adding 10 to the position of the farthest path with power within -20 dB of the main path as the channel length.

Images