KR20090043641A - Wireless Communication System Estimating Channel Using Efficient Training Sequence and Its Simple Inverse - Google Patents

Abstract

The present invention relates to a wireless communication system for estimating a channel using an efficient training sequence and a simple inverse thereof. The transmitting apparatus of the wireless communication system periodically transmits the training sequence signal between data signals to be transmitted with a training sequence set in advance for channel estimation. Here, the predetermined training sequence s has a length of 2n-1 and an alternating coded training sequence with the same sign for the n-2, n-1, and nth trains or the alternating code for the n-1, n, n + 1th trains. A coded training sequence or an alternating coded training sequence having the same n, n + 1, n + 2th code, is configured such that an inverse of the Toeplitz matrix with respect to the alternate coded training sequence exists. The receiving device of the wireless communication system has an inverse of the Toplit matrix with respect to the training sequence of the transmitting device. When the signal is received from the transmitter, the receiver obtains the product of the inverse of the training sequence and the received signal, calculates an impulse response to the received signal, and estimates the channel. According to the estimated impulse response of the channel, it is possible to recover and decode the original signal related to the digital communication system such as digital television broadcasting, mobile communication, portable Internet communication signal, etc. in a subsequent demodulation process.

Wireless communication, training sequence, inverse matrix, channel estimation

Description

Wireless communication system for estimating channels using effective training sequence and simple inverse matrix}

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a wireless communication system, and in particular, by using a short training sequence of binary signals made to enable efficient channel estimation, by flattening the frequency spectrum of a channel and defining three or five kinds of channel lengths. The present invention relates to a wireless communication system capable of improving channel estimation performance by accurately performing channel estimation on a received signal with a small amount of computation through a simple inverse matrix composed of elements.

In a wireless communication system such as digital television broadcasting, mobile communication, and the portable Internet, a method of using a training sequence to estimate an impulse response of a channel has become common. When estimating a channel with a training sequence, channel estimation using a decision direct technique is common when the impulse response of a channel changes slowly. However, if a sudden channel change occurs, such as a mobile receiving environment, channel estimation using this direct decision technique causes a significant performance degradation in the receiver.

In real systems, channel estimation mainly uses a method that satisfies the requirement of minimizing mean square error (MSE). The most commonly used are Least-Mean-Square (LMS) and Recursive-Least-Squares (RLS) algorithms. Although the performance of the RLS algorithm is excellent, the LMS algorithm is preferred because the calculation of the LMS is less than that of the RLS. A fast Kalman algorithm has been developed to reduce the computational complexity of the RLS algorithm, but this also requires more computation than the LMS.

These channel estimation methods require a tradeoff between computation and performance required for channel estimation. Theoretically, the channel can be estimated simply by using the product of the inverse of the training sequence of the Toeplitz matrix and the received signal. However, various channel estimation methods have been devised because of the large amount of computation and the possibility that the inverse does not exist. If the complement of the problem of the inverse method using the product of the training matrix inverse matrix and the received signal, accurate channel estimation will be possible without using complicated algorithms such as LMS or RLS algorithm.

Accordingly, an object of the present invention is to solve the above-described problem, and an object of the present invention is to accurately estimate a channel estimate for a received signal using an inverse matrix of a previously trained and stored training sequence. The present invention provides a wireless communication system that can be easily and simply performed.

A feature of the present invention for achieving the above technical problem is a wireless communication system for providing communication between a transmitter and a receiver using a communication channel of length n (where the length normalizes the interval of one symbol of the transmission signal to 1). As for

The transmitter has a training sequence preset for channel estimation, and periodically transmits the training sequence signal between data signals to be transmitted.

Wherein the training sequence has a length of (2n−1), and there is an inverse of the toplet matrix of the training sequence, and the inverse is determined by the length (n) of the communication channel,

The receiver stores an inverse matrix for a preset topless matrix of the training sequence, and the receiver calculates a channel impulse response using a product of the inverse matrix and a received signal received from the transmitter, and uses a channel impulse response. It is characterized by estimating.

The training sequence according to the above-described feature,

Alternate code training sequence containing 2n-1 elements where n is odd and the three centers s _n-2 , s _n-1 , and s _n are the same sign

이거나,

Or

Alternate code training sequence containing 2n-1 elements where n is even and the three centers s _n-2 , s _n-1 , and s _n are the same sign

이거나,

Or

Alternate code training sequence containing 2n-1 elements where n is odd and three s _n , s _{n + 1} , and s _{n + 2 in the center} are the same sign

이거나,

Or

Alternate code training sequence containing 2n-1 elements where n is even and the three centers s _n , s _{n + 1} , and s _{n + 2} have the same sign

이거나,

Or

Alternate code training sequence containing 2n-1 elements where n is odd and three s _n-1 , s _n , and s _{n + 1 in the center} are the same sign

이거나,

Or

Alternate code training sequence containing 2n-1 elements where n is even and the three centers s _n-1 , s _n , and s _{n + 1} are the same sign

인 것이 바람직하다.

Is preferably.

The signal transmitted and received in the wireless communication system having the above-mentioned characteristics is preferably one of a digital television broadcast, a mobile communication, and a portable internet communication signal.

According to the wireless communication system according to the present invention, by using a short training sequence having a length of (2n-1), by calculating a simple inverse matrix greatly reduced the amount of calculation, and storing the calculated inverse matrix in the receiving device, the stored inverse matrix and the received signal By multiplying with, it is possible to accurately and simply perform channel estimation on the received signal.

In addition, according to the wireless communication system according to the present invention, by using a short training sequence of binary signals made to enable efficient channel estimation, three or five kinds defined by the channel length while flattening the frequency spectrum of the channel It is possible to improve channel estimation performance through a simple inverse of elements.

In addition, according to the wireless communication system according to the present invention, when comparing the MSE between the channel estimated in the Brazilian channel and the TU6 channel and the signal passing through the actual channel, the proposed training sequence than when using the training sequence used for 8-VSB We can reduce the amount of computation during channel estimation, get 1dB to 9dB of channel prediction gain with 2n-1 training sequences, and obtain excellent channel estimation characteristics over the entire frequency.

In addition, the wireless communication system according to the present invention can be utilized in a wireless communication system such as digital television broadcasting, mobile communication, mobile Internet, and the like to improve reception performance of a wireless communication signal.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings and the contents described in the accompanying drawings, but the present invention is not limited or limited to the embodiments. Like reference numerals in the drawings denote like elements.

In the communication system, channel estimation is performed in the training sequence. The channel estimation compensates for the distorted signal and finds the actual transmitted signal. If the received signal is y, the channel impulse response is h, the training sequence s is a matrix S, and the white noise (additive white Gaussian noise, AWGN) is w, the received signal is expressed as shown in [Equation 1].

[Equation 1]

,

, S는 길이가 2n-1(n은 자연수)인 훈련열

로 만들어진 크기가 nㅧ n인 토플리츠(Toeplitz) 행렬, 그리고,

이며, T는 행렬전치이다. 각 벡터의 아래 첨자는 신호의 표본 시각을 나타낸다. 통신 시스템에 주로 사용하는 훈 련열 s는 1과 -1로 구성되기 때문에 이의 토플리츠 행렬 S는 이진 행렬이 되고 길이 n인 채널을 추정하기 위한 훈련열의 길이는 2n-1이 된다. here,

,

, S is a training train of length 2n-1 (n is a natural number)

A Toeplitz matrix of size n ㅧ n,

Where T is the matrix transpose. The subscript in each vector represents the sample time of the signal. Since the training sequence s mainly used in a communication system is composed of 1 and -1, its topless matrix S is a binary matrix and the length of the training sequence for estimating a channel of length n is 2n-1.

In order to obtain the channel impulse response, Equation 1 ignores AWGN and takes S ⁻¹ on both sides, and the approximate estimated channel ┎ is shown in Equation 2.

[Equation 2]

이다. [수학식 2]를 이용하여 채널을 추정할 경우, 일반적인 토플리츠 행렬의 역행렬을 얻기 위한 계산량은 O(n²)이다. 더구나 1과 -1로 구성된 의 S의 원소 배열에 따라 역행렬이 존재하지 않을 수 있기 때문에 실제 통신 시스템에 사용하기는 부적합하다.here

to be. In the case of estimating a channel using Equation 2, the amount of calculation for obtaining the inverse of the general topless matrix is O (n ² ). In addition, since the inverse may not exist depending on the element array of S consisting of 1 and -1, it is not suitable for use in a real communication system.

However, when S is an orthogonal matrix in which each row and column are orthogonal, S can be expressed as shown in [Equation 3] because the inverse matrix is equal to the transpose of S.

[Equation 3]

Where I represents the unit matrix. However, when n≥8, a matrix in which each row and column are orthogonal can not be obtained from a training column composed of 1 and -1. When using an orthogonal matrix, a matrix composed of training columns is constructed as shown in [Equation 4]. The number of elements required is n ² .

[Equation 4]

In the case of using such an orthogonal training sequence, channel estimation is possible without calculating an inverse matrix, but a long training sequence is required. In addition, in the case of ideal training trains, the frequency response should be flat. However, when using the orthogonal training train as described above, it is difficult to satisfy the flat frequency response characteristic as shown in the experimental results of FIGS. 10 to 20.

FIG. 1 is a flowchart sequentially illustrating a calculation process of obtaining an inverse matrix for the topples matrix of a training sequence in which only one element of a training sequence is -1 and all remaining ones are 1. Hereinafter, a process of calculating an inverse of the topple matrix of the training sequence will be described with reference to FIG. 1.

First, if we look at the training sequence where only one element of the training sequence is -1 and all the others are 1, it is easy to see that the inverse matrix exists only when the nth or n ㅁ first is -1, and there is no need for a separate operation to obtain the inverse matrix. Can be. If n-1 is -1, they are symmetrical, so let's look at the inverse of n-1 and n -1.

If you define the first training sequence of 2n-1 elements (where n is -1) as shown in [Equation 5], the diagonal elements are all -1 and the remaining elements are 1 6] can generate a binary topless matrix of n ㅧ n as shown in (S110).

[Equation 5]

[Equation 6]

In this case, the inverse of [Equation 6] can be expressed as shown in [Equation 7]. Here, det represents a determinant, and when S _{i, j} is an element of the i-th row and the j-th column of S, D _{i, j} is an argument matrix for S _{i, j} .

[Equation 7]

In order to calculate the inverse matrix as described above, first, the argument matrix D _{i, j} may be generated (S120), and then the determinant det D _{i, j} may be calculated (S130).

To this end, first, D _1,1 is expressed as shown in [Equation 8].

[Equation 8]

The elements constituting D _1,1 are all 1 except for the diagonal component. The matrix in which the first and last rows of D _1,1 are swapped can be expressed as shown in [Equation 9].

[Equation 9]

In Equation 9, if Gaussian is erased for the first column, Equation 10 is used.

[Equation 10]

In a similar manner, the matrix subjected to Gaussian elimination in Equation 8 is shown in Equation 11, and its determinant det D _1,1 becomes (n-3) 2 ^n-2 .

[Equation 11]

Similarly, generating the matrix D _1,2 for S _1,2 and applying the Gaussian elimination method as above gives the determinant detD _1,2 = (-2) ^n-2 . Using a similar method, detD _1,3 = 2 (-2) ^n-2-1 , ... detD _{1, j} = 2 ^(j-2) (-2) ^nj . Therefore, the determinant detS for S is expressed by Equation 12.

[Equation 12]

If i = j for each argument matrix, then detD _{i, j} = (n-3) 2 ^n-2 as in Equation 13 _, and if i ≠ j _, detD _{i, j} = as in Equation 14 2 ^(j-2) 2 ^nj

[Equation 13]

detD _{i, j} = (n-3) 2 ^n-2 (i = j)

[Equation 14]

detD _{i, j} = 2 ^(j-2) 2 ^n-2 (i ≠ j)

Accordingly, since detS ≠ 0 from Equation 12, an inverse of S exists. The inverse S- ¹ of S may be calculated by the formula [Equation 15] on the basis of the above factors and determinants (S140).

[Equation 15]

On the other hand, similarly, if the second training sequence (n-1th is -1) consisting of 2n-1 elements in length is defined as [Equation 16], all components immediately above the diagonal become -1. The remaining elements can generate a binary topless matrix of n ㅧ n such that Equation 17 is 1.

[Equation 16]

[Equation 17]

With respect to this, if the inverse matrix is obtained according to the above processes (S120 to S140), Equation 18 is obtained.

Equation 18

In this case, when n≥3, the inverted matrix always includes the inverse matrix formed from the above training sequence, and since the inverse matrix of S is determined by n from [Equation 18], calculation of the inverse matrix requires only simple division. Therefore, if we construct the matrix as shown in [Equation 4] according to the proposed training sequence as shown in [Equation 5] and [Equation 16], the amount of calculation is greatly reduced when the channel estimation using the inverse matrix and the orthogonal training sequence is used. The number of trains can be reduced from n ² to 2n-1.

As shown in Equation 5 and Equation 16, the proposed training sequence has a number of -1s and the remainder of all 1s, so that the frequency spectrum may include many specific frequency components. To achieve a flat spectrum across the band, such as an ideal impulse, 1 and -1 must alternately occur with equal probability. That is, the training train should have the following properties.

[P1] The numbers of 1 and -1 must be similar.

[P2] 1 and -1 should occur alternately if possible.

If the frequency spectrum component of the channel uses a non-flat training sequence, an error occurs in channel estimation. The orthogonal training sequence can be used to estimate the channel to equalize the distribution of elements 1 and -1 in the training sequence, but the length of the training sequence should be long as described above.

Hereinafter, the training sequence for improving the performance of channel estimation by configuring the number of -1 and 1 similarly to the proposed training sequence as shown in [Equation 5] and [Equation 16] will be described.

The case where n is even is very similar to that of odd, so let n be odd. The training sequence starts with 1 and the code shift training sequence in which the signs are alternated is shown in [Equation 19].

[Equation 19]

However, there is no inverse of the Topples matrix in this training sequence. As before, the only code shift training sequence with inverse is the following proposed training sequence.

The first proposed code shift training sequence uses s _n as -1 instead of 1 in the training sequence consisting of 1 and -1 repeatedly as shown in [Equation 20].

[Equation 20]

S is expressed as Equation 21 when the binary topless matrix is formed from the training sequence of Equation 20.

[Equation 21]

This is equivalent to multiplying all odd rows and columns of Equation 6 [Equation 6] by -1 in the first proposed training column above, so the inverse of Equation 21 is shown in Equation 22.

[Equation 22]

In the second proposed training sequence, s _n- 1 is used as 1 instead of -1 in the training sequence consisting of 1 and -1 repeatedly, as shown in [Equation 23].

[Equation 23]

When the binary topless matrix is formed from the training sequence of Equation 23, Equation 24 is expressed.

[Equation 24]

This is equivalent to multiplying all the even rows and columns of the topless matrix [Equation 17] of the training column proposed above by -1 by -1, so that the inverse of Equation 24 is shown in Equation 25.

[Equation 25]

If we apply ₁ to s _{n + 1} , we can obtain the inverse of the transpose of the result of the inverse matrix applied to s _n-1 .

The following is a summary of the proposed training sequence and the inverse of the topless matrix. If the training sequence starts with -1 instead of 1, then the training sequence is the same as the next training sequence multiplied by -1. In other words,

1. The training sequence s _i of an alternating coded training sequence containing 2n-1 elements whose n is odd and the three centers s _n-2 , s _n-1 , s _n are the same sign and the inverse S of the toplet matrix _ij ^-1 is the same as [Equation 26].

[Equation 26]

2. The training sequence s _i of an alternating coded training sequence containing 2n-1 elements with n being even and the three central s _n-2 , s _n-1 , s _n having the same sign and the inverse S of its topple matrix _ij ^-1 is the same as [Equation 27].

[Equation 27]

3. The alternating code training sequence s _i containing 2n-1 elements whose n is odd and the three centers s _n , s _{n + 1} , and s _{n + 2} have the same sign and the inverse S _ij ^{− 1} is the same as [Equation 28].

[Equation 28]

4. The alternating code training sequence s _i containing 2n-1 elements whose n is even and the three centers s _n , s _{n + 1} , and s _{n + 2} have the same sign and the inverse of the toplet matrix S _ij ^{− 1} is the same as [Equation 29].

[Equation 29]

5. Alternate code training sequence s _i containing 2n-1 elements whose n is odd and the three centers s _n-1 , s _n , s _{n + 1} have the same sign, and the inverse S _ij ^{− 1} is the same as [Equation 30].

Equation 30

6. Alternate code training sequence s _i containing 2n-1 elements whose n is even and the three centers s _n-1 , s _n , s _{n + 1} have the same sign and the inverse S _ij ^{− 1} is the same as [Equation 31].

Equation 31

As mentioned so far, by using the inverse of the toplit matrix for the alternating code training sequence of length (2n-1) (Equation 26] to Equation 31], as shown in Equation 2 The impulse response of the channel estimated from the received signal y can be calculated. Accordingly, by reconstructing and decoding original signals related to digital television broadcasting, mobile communication, and portable Internet communication signals in a subsequent demodulation process, noise may be removed to a user to provide a communication service having improved quality.

In order to compare the performance of channel estimation with respect to the various training sequences according to the present invention, a Brazilian channel which is widely used for verifying digital TV reception performance is used. We also compared the performance of the TU6 channel, which is mainly used for verifying communication systems in mobile environments.

The Brazilian channel is divided into five channel models. The outdoor antenna reception models, Channel A and Channel B, are relatively good channel environments, the indoor and outdoor intermediate reception models, Channel C, the indoor antenna reception models, Channel D, and single frequency network (SFN) models. In channel E represents a very poor channel environment.

[Table 1] shows the magnitude, relative magnitude [dB], and delay values of the main path and the multipath according to the Brazilian channel model. Channel A represents a magnitude and delay value when the multipath is received by the outdoor antenna having a smaller size and a shorter delay than the main path. The largest multipath signal is less than -13dB compared to the main path and has a delay of less than 6μs. Channel B is also a poorer channel environment than channel A, because the first multipath is slightly larger than the main path and has a long delay of 12.7 μs.

Channel C, an indoor and outdoor mid-point reception environment, is similar in size to the main path, consists of multipath with low delay, and pre-ghost is also present. Channel D, which is an indoor antenna reception environment, has many pre-ghost signals, and a multipath similar in size to the main path is received with a delay value similar to that of channel A. Channel E, which is an SFN channel environment, is an environment in which three signals of the same size receive only delay values, and are received at three transmitter intermediate points. The magnitude and delay values of the Brazilian channel are shown in Table 1, and the impulse response for each channel when transmitting a signal at 19.39 Mbps, which is the ATSC 8-VSB transmission rate, is shown in FIGS. 2 to 6.

TABLE 1

TU6 채널은 각 다중 경로가 도플러 효과를 갖도록 모델링 된 채널로서 이동 수신 환경을 나타낼 때 주로 사용된다. 도 7은 19.39Mbps로 신호를 전송할 때의 TU6 채널의 임펄스 응답을 나타낸다.

The TU6 channel is a channel modeled so that each multipath has a Doppler effect and is mainly used to represent a mobile reception environment. 7 shows an impulse response of a TU6 channel when transmitting a signal at 19.39 Mbps.

To compare with the actual system, we compared the performance of the training sequence used in the 8-VSB (8-level vestigial sideband modulation) system with the proposed training sequence. The training sequence of 8-VSB is shown in FIG. 8. The training sequence of 8-VSB consists of PN511, bits from PN63 generator, VSB mode bits, 104 bits of spare space, and uses PN511 training sequence mainly for estimation of radio channel.

Since the maximum channel delay value is different for each channel during the experiment, the channel length n is set to 128 samples, which is the maximum length of the channel, to represent the delay values of all the channels. 9 shows the frequency spectrum of S composed of the Brazilian A channel and each training sequence. When the diagonal component of S is a training sequence matrix of -1 (Diagonal -1 Toeplitz), the frequency characteristic is large and the gain of the DC component is flat. When S is an alternating coded topless matrix (Alternative Toeplitz), the gain of a specific high frequency region is large and is flat in the remaining frequency regions. The channel estimation results are multiplied by the spectrum of the training sequence, so a good performance can be obtained when the frequency spectrum has a flat area. When the matrix S is composed of an orthogonal training sequence and an 8-VSB training sequence (PN511), the frequency response varies with each frequency domain, but does not show very large or small gain in a specific frequency domain. Therefore, it can be considered to have a constant value on average.

In order to indicate the performance of channel estimation, the mean square error (MSE) between the estimated channel and a symbol passing through a given channel is compared with FIG. 10. The frequency spectrum of the Brazilian A channel has a high low frequency component and a low high frequency component. Therefore, if the diagonal component of S is -1, the channel estimation performance is 4.5dB higher than that of the 8-VSB training train and the orthogonal training train. In the case where S, the training sequence proposed in the present invention, is an alternating code topple matrix, the gain of the high frequency component of the channel is large, so that the gain of about 2dB is obtained compared with the use of the training sequence and the orthogonal training sequence used for 8-VSB. Get

Brazil B, C, and D channels (Brazil B, Brazil C, Brazil D) have a spectral distribution with a large direct current component and a small high frequency component like the Brazilian A channel. Therefore, the MSE performance for each training train is similar to that of the Brazilian A channel. Spectra of the B, C, and D channels of Brazil and the spectra of each training sequence are shown in FIGS. 11, 13, and 15, respectively, and channel estimation performance of the Brazilian B, C, and D channels of each training sequence is shown in FIGS. 16 is shown.

In the case of the Brazilian E channel, unlike the Brazilian channels A, B, C, and D, the channel spectrum has the same large gain in the low frequency band and the high frequency band, so that the channel estimation performance has the diagonal component of S equal to -1. For the alternate sign topple matrix, the performance is similar, and the performance is about 4.5dB lower than that of the 8-VSB training train and the orthogonal training train. Spectrum of the Brazilian E channel and each training train is shown in FIG. 17, and MSE according to signal-to-noise ratio (SNR) is shown in FIG. 18.

The TU6 channel has a large direct current component while the smallest high frequency region. Therefore, in the case of constructing the proposed alternating code topless matrix, it can be seen that the performance of channel estimation is 9dB higher than that of the 8-VSB training sequence. The spectrum of the TU6 channel and each training sequence is shown in FIG. 19, and the MSE according to SNR is shown in FIG. 20.

The use of binary alternating code topless matrices has all the advantages of a binary symmetric topless matrix with a diagonal component of -1, while also providing good MSE performance. In other words, it is a training sequence structure that does not require inverse matrix calculation and improves channel estimation performance with a short training sequence. Table 2 shows the MSE for each channel compared with the PN511 training sequence. Here, the-sign indicates performance deterioration.

TABLE 2

A transmission apparatus for wireless communication applying the channel estimation method according to the present invention as described above has a training sequence preset for channel estimation, and modulates the signal to be transmitted by using the training sequence and then modulates the transmission. do. Here, the predetermined training sequence s has a length of 2n-1 and an alternating coded training sequence with the same sign for the n-2, n-1, and nth trains or the alternating code for the n-1, n, n + 1th trains. As a code training sequence or an alternating code training sequence having the same number as the n, n + 1, n + 2 th ([Equation 26] to [31]), the inverse of the Toplet matrix with respect to the alternating code training sequence Configured to exist.

The receiving apparatus for wireless communication, to which the channel estimation method according to the present invention is applied, includes an inverse of the Toplet matrix with respect to the training sequence of the transmitting apparatus. When the signal is received from the transmitter, the receiver obtains the product of the inverse of the training sequence and the received signal, and calculates an impulse response 에 for the received signal y as shown in [Equation 2] to estimate the channel. According to the estimated impulse response of the channel, it is possible to recover and decode the original signal related to digital television broadcasting, mobile communication, portable Internet communication signal, etc. in a subsequent demodulation process.

As described above, the training sequence proposed in the present invention can make the frequency spectrum flat when the training sequence is composed of binary signals, and the inverse matrix is composed of only three kinds of elements, and the element value can be defined by the channel length. Therefore, the number of operations can be greatly reduced in channel estimation, and the channel estimation can be accurately performed in a short training sequence. When comparing the MSE between the channel estimated in the Brazilian channel and the TU6 channel and the signal passing through the real channel, the proposed training sequence can be used to reduce the amount of computation when estimating the channel rather than using the training sequence used for 8-VSB. With two training sequences, we can obtain channel prediction gains of 1dB to 9dB.

As described above, the present invention has been described by way of limited embodiments and drawings, but the present invention is not limited to the above embodiments, and those skilled in the art to which the present invention pertains various modifications and variations from such descriptions. This is possible. Therefore, the scope of the present invention should not be limited to the described embodiments, but should be determined not only by the claims below but also by the equivalents of the claims.

1 is a flowchart illustrating a process of calculating an inverse matrix for a topless matrix of an arbitrary training column.

2 is a diagram for explaining an impulse response of a Brazilian A channel.

3 is a diagram for describing an impulse response of a Brazilian B channel.

4 is a diagram for explaining an impulse response of a Brazilian C channel.

5 is a diagram for explaining an impulse response of a Brazilian D channel.

6 is a diagram for explaining an impulse response of a Brazilian E channel.

7 is a diagram for describing an impulse response of a TU6 channel.

8 is a diagram for explaining a training sequence structure of 8-VSB.

FIG. 9 is a diagram for describing each training sequence and power spectrum density (PSD) of the Brazilian A channel.

FIG. 10 is a diagram for describing a mean square error (MSE) in a Brazilian A channel. FIG.

FIG. 11 is a diagram for describing each training sequence and power spectrum density (PSD) of the Brazilian B channel.

12 is a diagram for explaining a mean square error (MSE) in a Brazilian B channel.

FIG. 13 is a diagram for describing each training sequence and power spectrum density (PSD) of the Brazilian C channel.

14 is a diagram for describing a mean square error (MSE) in a Brazilian C channel.

FIG. 15 is a diagram for describing each training sequence and power spectrum density (PSD) of the Brazilian D channel.

FIG. 16 is a diagram to describe Mean Square Error (MSE) in a Brazilian D channel.

FIG. 17 is a diagram for describing each training sequence and power spectrum density (PSD) of the Brazilian E channel.

18 is a diagram for explaining a mean square error (MSE) in a Brazilian E channel.

FIG. 19 is a diagram for describing Power Spectrum Density (PSD) of each training sequence and TU6 channel.

20 is a diagram for describing a mean square error (MSE) in a TU6 channel.

Claims (5)

A wireless communication system providing communication between a transmitter and a receiver by using a communication channel of length n, The transmitter has a training sequence preset for channel estimation, and periodically transmits the training sequence signal between data signals to be transmitted, wherein the training sequence has a length of (2n-1) and the training sequence. There is an inverse of the topple matrix for the column and the inverse is determined by the length (n) of the communication channel, The receiver has an inverse matrix for the topple matrix of the training sequence, which is calculated and stored in advance, and the receiver uses the product of the inverse matrix and the received signal received from the transmitter, using the inverse of the topple matrix of the training sequence. Calculating a channel impulse response, and estimating a channel using the channel impulse response. The method of claim 1, wherein the training train, Alternate code training sequence containing 2n-1 elements where n is odd and the three centers s _n-2 , s _n-1 , and s _n are the same sign

Or an alternating coded training sequence comprising 2n-1 elements in which n is even and the three centers s _n-2 , s _n-1 , and s _n are the same sign

A wireless communication system, characterized in that. The method of claim 1, wherein the training train, Alternate code training sequence containing 2n-1 elements where n is odd and three s _n , s _{n + 1} , and s _{n + 2 in the center} are the same sign

Or an alternating coded training sequence comprising 2n-1 elements, wherein n is even and three of the centers s _n , s _{n + 1} , and s _{n + 2} have the same sign

A wireless communication system, characterized in that. The method of claim 1, wherein the training train, Alternate code training sequence containing 2n-1 elements where n is odd and three s _n-1 , s _n , and s _{n + 1 in the center} are the same sign

Or an alternate code training sequence containing 2n-1 elements in which n is even and three s _n-1 , s _n , s _{n + 1 in the center} are of the same sign

A wireless communication system, characterized in that. The wireless communication system according to claim 1, wherein the received received signal is one of a digital television broadcast, a mobile communication, and a portable internet communication signal.

Images