Reduced complexity sliding window based equalizer

Equalizer based on reduced complexity sliding window

Technical field

The present invention generally relates to wireless communication systems. In particular, the present invention relates to data detection in this system.

BACKGROUND OF THE INVENTION Due to the increased demand for improved receiver performance, many advanced receivers use zero forcing (ZF) block linear equalizers and minimum mean square error (MMSE) equalizers.

In these two methods, the received signal is typically modeled according to Equation 1.

r=Hd+n Equation 1r is the received vector, including the sample of the received signal. H is the channel response matrix. D series data vector. In extended spectrum systems, such as code division multiple access (CDMA) systems, d is the extended data vector. In the CDMA system, the data of each individual cipher is generated by using the cipher to non-extend the data vector d. n is the noise vector.

In a ZF block linear equalizer, the data vector is estimated, for example, according to Equation 2.

d=(H)-1r Equation 2(·)H is a complex conjugate shift term (or Hermetian) operation. In an MMSE block linear equalizer, the data vector is estimated, for example, according to Equation 3.

d=(HHH+σ2I)-1r Equation 3 In wireless channel experience multi-path value-added, in order to use these methods to accurately detect data, a very large number of received samples need to be used. One method to reduce complexity is a sliding window method. In the sliding window method, a predetermined window and channel response system for receiving samples are used in data detection. After the initial test, the window is slid down to a sample next door. This process continues until the communication stops.

By not using an extremely large number of samples, an error is introduced into the data detection. The error is most significant at the beginning and end of the window, where the effectively truncated part of the extremely large sequence has the greatest effect. One way to reduce these errors is to use a large window size and truncate the results at the beginning and end of the window. The truncated part of the window is determined by the previous or next window. This method has considerable complexity. The large window size results in a large size of the matrix and vector used for data estimation. In addition, this method detects that the data is not computationally valid at the beginning and end of the window and then discards the data.

Accordingly, we want to have alternative data detection methods.

Summary of the invention

The data estimation is executed in a wireless communication system. A receiving vector is generated. With the data used to estimate a desired part of the received vector, a past, a center and a future part of a channel estimation matrix are determined. The past part and the part of the received signal have priority over the desired part of the data. The future part is associated with a part of the received vector after the desired part of the data and the central part is associated with a part of the desired part of the received vector. The desired part of the data is the detection data that is estimated to have no effective truncation. Estimate the desired part of the data using a minimum mean square error algorithm with the input of the central part of the channel estimation matrix and a part of the received vector. The past and future parts of the channel estimation matrix are used to correct the coefficients in the minimum mean square error algorithm.

Description of the drawings

Figure 1 is an icon illustrating a response matrix with channels.

Figure 2 is an icon illustrating a central part of the channel response matrix.

Figure 3 is an icon illustrating a possible division of a data vector window.

Figure 4 is an icon illustrating a divided signal mode.

Fig. 5 is a flow chart of a past calibration system used in sliding window data detection.

Figure 6 is a receiver using a sliding window data detection method described in the past calibration system.

Fig. 7 is a flow chart of using a noise automatic correlation correction coefficient for sliding window data detection.

Figure 8 is a sliding window data detection in which a receiver uses a noise automatic correlation correction coefficient.

detailed description

Thereafter, a wireless transmission/reception unit (WTRU) includes but is not limited to a user device, a mobile base, a fixed or mobile wireless telephone unit, a pager, or any type of device that can operate in a wireless environment. When referring to this, a base station includes but is not limited to a Node-B, location controller, access point or any other type of interface device in a wireless environment.

Although the reduced complexity sliding window equalizer is described as related to a better wireless code division multiple access communication system, such as CDMA2000 and Global Mobile Terrestrial System (UMTS) Frequency Division Duplex (FDD), Time Division Duplex (TDD) Mode and Time Division Synchronous CDMA (TD-SCDMA), which can be used in a variety of communication systems and, in particular, different wireless communication systems. In a wireless communication system, it can be used to transmit and receive from a base station via a WTRU, from one or more WTRUs via a base station, or receive from another WTRU via a WTRU, such as an ad hoc mode operation in.

The following text describes a sliding window reduction method based on an equalizer using a better MMSE algorithm. However, other algorithms can be used, such as a zero forcing algorithm. h(·) is the impulse response of one channel. d(k) is the kth transmission sample, which is generated by unfolding a symbol and using an unfolding code. It can also be the sum of the chips, which is generated by expanding a set of symbols and using a set of ciphers, such as holding ciphers. r(·) is the received signal. The model of the system can be expressed in accordance with Equation 4.

r(t)=Σk=-d(k)h(t-kTc)+n(t),-<t<])>Equation 4n(t) is additional The sum of noise and interference (intra-cell and inter-cell). For simplicity, the following text is described assuming that chip rate sampling is used in the receiver, although other sampling rates may be used, such as a multiple of the chip rate. The sampled received signal can be represented as in accordance with Equation 5.

r(j)=Σk=-d(k)h(jk)+n(j)=Σk=-d(jk)h(k)+n(j) ,j{...,-2,-1,0,1,2,...}]]>Equation 5Tc system is stopped to simplify notation.

Suppose h(k) has a limited support and is time-invariant. This means that in the separation time range, the index L has h(i)=0 and i<0 and iL. As a result, Equation 5 can be rewritten as Equation 6.

r(j)=Σk=0L-1h(k)d(jk)+n(j),j{...,-2,-1,0,1,2,...}]] >Equation 6 considers that the received signal has M received signals r(0),...,r(M-1), and the result is equation 7.

r=Hd+n where r=[r(0),...,r(M-1)]TCM,

d=[d(-L+1), d(-L+2),...,d(0), d(1),...,d(M-1)]TCM+L=1n=[n (0),...,n(M-1)]TCMThe part of the vector d in Equation 7 can be determined using an approximation equation. Assuming M>L and defining N=M-L+1, the vector d is in accordance with equation 8.

d=[d(-L+1), d(-L+2),..., d(-1), d(0), d(1),..., d(N-1), d(N) ,...,D(N+L-2)]TεCN+2L-2 Equation 8 The H matrix in Equation 7 is a banded matrix, which can be represented as the icon in Figure 1. In FIG. 1, each column in the shaded area represents a vector [h(L-1), h(L-2),..., h(1), h(0)], as shown in Equation 7.

Instead of estimating all the components in d, only the middle N components of d are estimated. The system is the middle N component as shown in accordance with equation 9.

d~=[d(0),...,d(N-1)]T]]>Equation 9 uses the same observation data for r, a linear relationship with r and The time is in accordance with Equation 10.

r=H~d~+n]]>Equation 10 matrix It can be represented as shown in Figure 2 or as in accordance with Equation 11.

As shown in Equation 11, the first L-1 and last L-1 components of r are not equal to the numbers on the right hand side of Equation 10. As a result, in the vector Components at both ends will be estimated less accurately than those closer to the center. Because of this characteristic, a sliding window method is preferably used to estimate transmission samples, such as chips.

In each of the kth steps of the sliding window method, a reliable number of received samples is maintained in r(k) with size N+L-1. They are used to estimate a set of transmitted data Use Equation 10 with size N. In vector After being estimated, only the estimated vector The middle part of the is used for further data processing, for example by reducing the stretch. The "lower part (or the later part of the time) is estimated again in the next step of the sliding window procedure, where r[k+1] has some of the components r[k] and some newly received samples, namely its It is an offset (sliding) way of r[k].

Although, preferably, the window size N and the sliding step size are design parameters, (based on the extension of the delay channel (L), the accuracy requirements of the data estimation and the method complexity limitations), the following text uses the equation 12 of the window size To illustrate, N=4Ns×SF equation 12SF is the extension coefficient. The typical window size is 5 to 20 times larger than the channel impulse response, although other sizes may be used.

The sliding size is based on the window size system of Equation 12, and is preferably 2Ns×SF. N {1, 2,...} is preferably left as a design parameter. In addition, in each sliding step, the estimated chip system sent to the lower stretcher is 2Ns×SF components. in the middle. This step is illustrated in Figure 3.

An algorithm for data detection uses an MMSE calculus with model error correction using a sliding window-based method and the system model of Equation 10.

Because of the approximation, the estimation of the data, such as the chip, has errors, in particular, at both ends of the data vector in each sliding step (start and end). In order to correct this error, the H matrix is divided into a block column matrix in Equation 7 as in accordance with Equation 13 (step 50).

H=[Hp|H~|Hf]]]>Equation 13 subscript symbol "p" stands for "past", and "f" stands for "future". The system is as in accordance with equation 10. Hp is as in equation 14.

Equation 14

Hf is as in equation 15.

The vector d system of Equation 15 is also divided into blocks as in Equation 16.

d=[dpT|d~T|dfT]T]]>Equation 16The system is the same as Equation 8 and the dp system is the same as Equation 17.

dp=[d(-L+1)d(-L+2)...d(-1)]TεCL-1

df=[d(N)d(N+1)...d(N+L-2)] TεCL-1 Equation 18 The original system model is then shown in Equation 19 and shown in FIG. 4.

r=Hpdp+H~d~+Hfdf+n]]>Equation 19-The method takes Equation 19 as the model system as Equation 20.

r~=H~d~+n~1]]> where r~=r-Hpdp]]> and n~1=Hfdf+n]]> Equation 20 uses an MMSE algorithm to estimate the data vector The system is shown in equation 21.

d~^=gdH~H(gdH~H~H+Σ1)-1r~^]]> Equation 21 In Equation 21, gd is the chip energy as in Equation 22.

E{d(i)d*(j)}=gdδij equation 22It is as equation 23.

r~^=r-Hpd^p]]>Equation 23Is based on the previous sliding window step The estimated part. 1 series The automatic incidence matrix of, that is, Σ1=E{n~1n~1H}.]]> If it is assumed that Hfdf and the n system are not related to each other, it is caused by Equation 24.

Σ1=gdHfHfH+E{nnH}]]>Equation 24The credibility of depends on the sliding window size (with respect to the channel delay across L) and the sliding step size.

This method is also described in connection with the flowchart of FIG. 5 and the preferred receiver component of FIG. 6, which can be implemented in a WTRU or base station. The circuit of FIG. 6 can be implemented on a single integrated circuit (IC), such as an application specific integrated circuit (ASIC), on multiple ICs, such as discrete components or a combination of an IC and discrete components.

A channel estimation device 20 processes the received vector r to generate a channel estimation matrix part Hp, And Hf, (step 50), a future noise automatic correlation device 24 determines a future noise automatic correlation coefficient, gdHfHfH, (step 52). A noise automatic correlation device 22 determines a noise automatic correlation coefficient, E{nnH}, (step 54). A sum calculator 26 calculates the sum of the two coefficients to generate Σ1, (step 56).

A past input correction device 28 takes the past part of the channel response matrix, Hp, and data vectorA part that was determined in the past to generate a past correction coefficient, (Step 58). A subtractor 30 subtracts the past correction coefficient from the received vector to generate a corrected received vector, (Step 60). An MMSE device 34 uses 1, as well as Decide to receive the central part of the data vector As in equation 21 (step 62). The next window is determined in the same way, using Part of as In the next window decision, (step 64). As explained in this method, only the data of the part of interest, It was decided to reduce the complexity involved in data detection and truncation of unwanted parts of the data vector.

In another method of data detection, only noise is corrected. In this step, the system model is as in Equation 25.

r=H~d~+n~2,]]>where n~2=Hpdp+Hfdf+n]]>Equation 25 uses an MMSE calculus, the estimated data vector The system is shown in equation 26.

d~^=gdH~H(gdH~H~H+Σ2)-1r]]>Equation 26 assumes that Hpdp, Hfdf and n are not related to each other, which is caused by Equation 27.

Σ2=gdHpHpH+gdHfHfH+E{nnH}]]>Equation 27 In order to reduce the complexity of using Equation 27 to solve Equation 26, a complete matrix multiplication system of HpHpH and HfHfH is not required, because only the upper and lower corners Hp and Hf are non-zero respectively, generally speaking.

This method is also described in connection with the flowchart of Figure 7 and the preferred receiver components of Figure 8, which can be implemented on a WTRU or base station. The circuit of FIG. 8 can be implemented in a single integrated circuit (IC), such as an application specific integrated circuit (ASIC), in multiple ICs, such as discrete components or a combination of ICs and discrete components.

A channel estimation device 36 processes the received vector to generate a channel estimation matrix part, Hp, And Hf, (step 70). A noise automatic correlation correction device 38 determines a noise automatic correlation correction coefficient, gdHpHpH+gdHfHfH, using the future and past parts of the channel response matrix (step 72). A noise automatic correlation device 40 determines a noise automatic correlation coefficient, E{nnH}, (step 74). A sum calculator 42 adds the noise automatic correlation correction coefficient to the noise automatic correlation coefficient to generate Σ2, (step 76). An MMSE device 44 uses the central part or channel response matrix Receive vector r and estimate the central part of the data vector (Step 78). One advantage of this method is that it does not need to use a feedback loop for the detection data. As a result, different sliding windows can be determined in parallel and discontinuously.