KR100885252B1 - Adaptive Filter Apparatus for Wireless Channel Estimation with Dual LMS Filter and Step Size Control Method Used Thereof - Google Patents

Abstract

Disclosed are an adaptive filter device for estimating a wireless channel and a step size varying method of an adaptive filter, which are used for detecting / estimating / removing a radio signal and time-controlling step size. An LMS filter apparatus for performing radio channel estimation with a main LMS adaptive filter, comprising: an auxiliary LMS filter for updating tap coefficients of the main LMS adaptive filter; Provide an LMS filter device. According to the present invention, it is possible to ensure fast convergence speed for multiple channels and to quickly detect coefficients of taps corresponding to new multipaths, while improving performance of reducing convergence errors with a small step size for the remaining taps. It is possible to provide an LMS filter having a.

Description

WIRELESS CHANNEL ESTIMATION ADAPTIVE FILTER COMPRISING DUAL LMS FILTER AND STEP SIZE CONTROL METHOD THEREOF}

The present invention relates to a wireless channel estimation system, and more particularly, used for detecting / estimating / removing a radio signal, and a step size control method of an adaptive filter and an adaptive filter for wireless channel estimation capable of time varying control of a step size. It is about.

Two major performance criteria for adaptive filters are adaptive speed and minimum mean square error (MMSE). Adaptive speed is particularly important in environments where channel interference changes quickly over time. The tracking capability of the adaptive filter depends on the rate at which the tap coefficients of the filter converge as the channel profile changes. The adaptive algorithm of the adaptive filter is known as least mean square (LMS), recursive least square (RLS), and the like. The RLS algorithm converges faster than the LSM algorithm but is more complicated to implement in the same number of taps.

In general, the LMS filter exhibits a problem of noise increase when the tap input is increased, and thus, a normalized LMS (NLMS) algorithm is generally applied to apply in a large input variation condition such as a wireless channel.

As described above, the NLMS filter is one of typical adaptive filters used for detecting / estimating / removing a radio signal. Examples of the NLMS filter include a detection and cancellation filter for a wireless feedback signal of an oscillation cancellation repeater.

An RF repeater for mobile communication is composed of a receiving antenna, an amplifier, and a transmitting antenna, and receives and amplifies a weak signal and retransmits the amplified signal. As described above, when the wireless repeater amplifying the same frequency signal does not sufficiently secure the separation distance between the transmitting and receiving antennas at the time of installation, a feedback loop is formed in which the amplified signal amplified and output through the amplifier is applied to the input of the wireless repeater. An oscillation phenomenon occurs in which the feedback signal is repeatedly amplified through the amplifier.

In more detail, the signal transmitted from the base station system to the repeater is input to the repeater receiving antenna, and the signal is amplified by a gain G by an amplifier and radiated by the transmitting antenna. Here, a part of the signal radiated by the transmitting antenna is led back to the receiving antenna via a surrounding reflector or the like, and a feedback loop is formed which is amplified by the repeater gain G. The path loss of the feedback signal from the transmitting antenna to the receiving antenna is called isolation. If the isolation is not more than a certain value compared to the repeater gain G, oscillation by the feedback loop occurs.

Accordingly, in order to solve such a problem, an interference cancellation system repeater (ICS) that detects / removes a feedback signal including a radio channel characteristic that causes oscillation is being operated.

1 is a diagram schematically showing a conventional NLMS filter device for a conventional interference cancellation repeater. FIG. 1 shows that a feedback channel 20 is introduced between a receiving end and a transmitting end of an interference cancellation repeater in a signal path.

1 is a step size parameter that most influences the convergence speed and error characteristics of the NLMS filter. Therefore, selecting the optimal step size directly affects the convergence characteristics and performance of the filter coefficients.

The tap coefficient update expression of the NLMS filter is well known and is as follows. When the filter input vector is x (n), the filter output vector is y (n), and the expected value of the input signal is d (n), the tap coefficient (w) update expression of the NLMS filter is expressed by the following equation.

w (n + 1) = w (n) -μ / ∥ x (n) ∥ x (n) e ^* (n)

Where w (n) = [w ₀ (n), w ₁ (n),... , w _M (n)] ^T and x (n)-[x (n), x (n-1),... , x (n-M + 1)] ^T , and e (n) = d (n) -y (n) = d (n) -∑ _k w _k (n) x (nk). In addition, k denotes a tap number of the filter, and μ denotes a step size for tap coefficient update.

In the related art, a fixed value of the step size is used in the NLMS filter. However, in the case of the fixed step size, when the μ is set small, the error after convergence is small but the convergence speed is slow. When the μ is set large, the convergence speed is fast but the error after the convergence is large.

Accordingly, efforts have been made to control the step size in accordance with the state of the filter tap coefficient by controlling the step size of the LMS filter variably. In order to estimate the fast changing time-varying radio fading channel, the convergence speed of the filter is increased by increasing the step size of the filter when the multipath channel is generated. To this end, conventionally, an appropriate step size control mechanism is added to estimate input signal power, channel change rate, and the like, and control the step size of all taps of the filter 10 to the same value over time. However, as described above, the increase in the step size can speed up the convergence, but still has the problem of causing an increase in the error after the convergence, thereby limiting the performance improvement effect due to the step size control.

As described above, in channel detection and estimation of a wireless multipath environment, the coefficient value of a specific tap corresponding to the multipath channel becomes overwhelming, and for the tap, a quick update of the coefficient according to the time-varying characteristics (the Doppler characteristics) of the wireless channel. While the count of taps that are not needed does not require a quick update.

Accordingly, an object of the present invention is to provide an LMS filter device suitable for time-varying radio fading channel estimation by reducing an estimation error despite fast radio channel change.

It is another object of the present invention to provide an LMS filter device capable of optimally maintaining the convergence speed and the error of the LMS filter.

Moreover, an object of this invention is to provide the step size control method suitable for the step size control of the above-mentioned LMS filter apparatus.

In order to achieve the above technical problem, the present invention provides a LMS filter apparatus including a main LMS adaptive filter to perform radio channel estimation, comprising: an auxiliary LMS filter for updating tap coefficients of the main LMS adaptive filter; An LMS filter apparatus for estimating a wireless channel is provided.

In the present invention, it is preferable that the main LMS filter and the auxiliary LMS filter have the same number of taps.

In addition, the step size for each tap coefficient of the auxiliary LMS filter is preferably fixed.

The present invention is characterized in that the tap-size step size of the corresponding main LMS filter is controlled based on the state of the tap-by-tap coefficient of the auxiliary LMS filter. In this case, the state of each tap coefficient which is the basis of step size control in the present invention may be, for example, the power of each tap coefficient of the auxiliary LMS filter.

In the present invention, the main LMS filter and the secondary LMS filter is preferably a normalized LMS filter.

According to another aspect of the present invention, there is provided a method for detecting a coefficient state of a plurality of taps of the filter with respect to a signal input to a first adaptive filter (auxiliary filter), and the detected first adaptive filter. Varying the step size of the corresponding tap of the second adaptive filter (main filter) on the basis of the coefficient state of each tap of the step; To provide.

In the present invention, the coefficient state of each tap of the first adaptive filter may be determined by the magnitude of the power of the complex coefficient of the tap of the first adaptive filter. In addition, the coefficient state of each tap of the first adaptive filter may be determined by an average value of power of complex coefficients of the tap of the first adaptive filter for a predetermined time. In this case, the step size varying step of the corresponding tap may include taps of the second adaptive filter corresponding to taps whose power or average power of a complex coefficient among taps of the first adaptive filter exceeds a preset first threshold value. To increase the step size of or reduce the step size of the tap of the second adaptive filter corresponding to the tap whose power or average power of the complex coefficient among the taps of the first adaptive filter is less than a second preset threshold. It can be made by.

In the present invention, the step size varying step of the corresponding tap is preferably performed through at least one transition state having a predetermined step size between the current step size and the final step size.

In addition, the step size in the step size variable step of the corresponding tap may be varied by Doppler frequency estimation.

In the present invention, the first and second adaptive filters may be LMS adaptive filters.

In addition, the step size of the first adaptive filter in the initialization step of the first and second adaptive filter is preferably larger than the step size of the second adaptive filter.

With the above configuration, the present invention sets the tap coefficient status for each tap coefficient of the LMS filter to control the step size. Accordingly, when there is a radio time-varying multipath to be estimated, a fast convergence speed is ensured by setting the corresponding tap to a large step size and updating the coefficients, and the remaining taps set a small step size and then update the coefficients to update the coefficients. Can be reduced. In this way, the performance of the LMS filter can be improved by setting and controlling the variable step size for each tap coefficient.

In addition, according to the present invention, the LMS filter passes through the noise state and the transition state to the transition to the search state so that the transition between the two states does not occur frequently. In addition, in the present invention, by adding a noise state that is determined to be not multipath and maintaining the LMS filter at a low step size, the error in a steady state can be kept small. As such, the present invention improves the performance of the filter in a multipath environment by providing hysteresis in transition of the count state of the tap so that transition between states does not occur frequently.

In addition, even when the multipath is detected, an optimal step size design is possible by estimating the Doppler frequency for each tap coefficient and controlling the step size based on this.

Hereinafter, exemplary embodiments of the present invention will be described with reference to the accompanying drawings.

2 is a view for explaining an example of a radio channel estimation system according to the present invention. 2 illustrates the use of an NLMS filter device in the present invention in an interference cancellation repeater, but the filter device and system of the present invention are not limited thereto and may be applicable to the estimation, detection or removal of a general radio channel.

2, the NLMS filter device of the present invention is composed of a main NLMS filter 100 and a secondary NMLS filter 120. The main NLMS filter 100 outputs the output vector y (n) by using a signal at a predetermined point of the repeater, for example, as the input signal x (n). The output of the main NLMS filter 100 is summed with the expected value d (n) of the input signal to obtain an error signal e (n). The tap coefficient of the main NLMS filter 100 is updated to reduce the error signal. The algorithm of Equation 1 described above may be used for minimizing an error signal, and this technique is well known to those skilled in the art.

In the present invention, the tap coefficient of the main NLMS filter 100 may be updated for each tap.

2 illustrates the use of a secondary NLMS filter as a preferred embodiment for coefficient-by-coefficient update of the NLMS filter. As shown, the auxiliary NLMS filter outputs the output vector y '(n) using the signal x' (n) of the repeater receiver as an input signal.

Preferably, the auxiliary NLMS filter 120 of the present invention has the same number of taps (not shown) as the main NLMS filter 120. Further preferably, the step size of the tap of the secondary NLMS is set to a relatively large value compared to the step size of the corresponding tap of the main NLMS filter. For example, if the step size initial value of the k-th tap of the main NLMS filter is 1/16384, the step size initial value of the k-th tap of the auxiliary NLMS filter may be set to 1/256, 1/512, 1/1024, or the like. have. In this way, by setting the step size of the tap of the auxiliary NLMS filter large, it is possible to quickly detect the amount of change in the filter tap coefficient of the newly generated and destroyed multipath position.

In addition, the step size of each tap of the auxiliary NLMS filter may be initially set to the same one value. Of course, the step size may be set to different values. For example, the step size μ _k of all taps of the auxiliary NLMS filter may initially be 1/256 or 1/512. In the present invention, it is preferable that the step size of each tap of the auxiliary NLMS filter is shifted step by step according to a multipath environment to improve the performance of the filter. This will be described later.

Further, in a time varying wireless environment, the step size of the main NLMS filter of the present invention is variable for each tap coefficient, while the step size of the auxiliary NLMS filter is preferably fixed.

In the present invention, the secondary NLMS filter is used to control the step size of the tap of the corresponding main NLMS filter with respect to its tap. Specifically, the power value | W _{s, k} | ² of each tap coefficient obtained in the auxiliary NLMS filter is used to control the step size of the corresponding main NLMS filter tap. A more specific method of controlling the step size of the main NLMS filter will be described later.

As described above, the filter device of the present invention includes a secondary NLMS filter, and continuously monitors the state of the tap coefficient according to the radio signal inputted to the secondary NLMS filter, so that the tap coefficient of the main NLMS filter corresponding to the multipath can be quickly increased. To ensure convergence.

Meanwhile, in the present invention, the time varying step size for each coefficient may be controlled by using a change amount of the filter tap coefficient in a multi-channel environment. Here, a method well known to those skilled in the art to which the present invention pertains, such as estimating the Doppler frequency for each coefficient using a tap coefficient change amount, may be used.

Hereinafter, a detailed method of controlling the step size of the main NLMS filter using the auxiliary NLMS filter 120 of the present invention will be described.

3 is a diagram schematically illustrating a step size state transition of a main NLMS filter using an auxiliary NLMS filter in the present invention.

Referring to FIG. 3, in the present invention, the count state of each tap of the main filter NLMS is divided into an initial state, a noise state, a transition state, and a search state. In the present invention, the state transition of the coefficients of each tap of the main filter is performed by comparing the power of the complex coefficients of each tap of the auxiliary NLMS filter with a preset threshold.

In the initial state, the step size of each tap of the main NLMS filter may be set to the same value. In addition, in the initial state, the step size of each tap of the auxiliary NLMS filter is set to an arbitrarily large value compared to the step size of the corresponding tap of the main NLMS filter. For example, the step size of all taps of the auxiliary NLMS filter 120 may be set to 1/256, and the step size of all taps of the main NLMS filter 100 may be set to a very small value (for example, 1/16384 or less). Can be.

In the initial state, the wireless channel estimation system measures the power of the complex coefficient of the tap of the auxiliary filter 120 and compares this value with a preset threshold to determine if there are multiple paths. For example, when the power of the complex coefficient power of the k-th tap (│W _{s , k} │ ² ) is greater than or equal to a preset threshold value (θ), the multipath corresponding to the k-th tap is sensed, and thus k of the main NLMS filter is detected. The first tap coefficient is shifted to the search state. In the search state, the step size of the corresponding tap of the main NLMS filter is changed to the value (μ _s ) of the search state. Since the multipath value fluctuates over time in the search state and the coefficients must be updated quickly, the step size μ _s of the corresponding tap of the main filter in the search state is set to a large value to quickly adapt to channel changes (e.g. 1/256). Or 1/512).

If the power of the plurality of coefficients of the k-th tap of the auxiliary filter 120 is less than the threshold value θ, it is determined that there is no multipath, and the state of the k-th tap coefficient of the main NLMS filter is a noise state. To transition to. In this case, since there is no multipath corresponding to the k th tap, μ _N of the corresponding tap of the main NLMS filter is set to a small value to reduce the error in the steady state (eg, μ _N = 1/16384). In the present invention, the threshold value [theta] used as a criterion for determining the transition from the initial state to the noise state can be appropriately set in consideration of the normal noise level and the complex coefficient power value in the multipath environment.

When the power of the tap coefficient of the secondary NLMS filter measured in the noise state is greater than or equal to the transition state threshold θ _NT , the corresponding tap coefficient of the main NLMS filter 120 transitions to a transition state. Here, the step size μ _T for determining the tap coefficient state of the transition state is preset to an appropriate value between μ _N of the noise state and μ _S of the search state. For example, μ _N may be set to 1/16384, μ _T to 1/4096, and μ _S to 1/256.

If the power value of k-th tap coefficient of the secondary filter measured at the transition state (transition state) is less than a predetermined threshold value (θ _TN) it is determined to be a noise state setting the step size of the k-th tap of the main filter with μ _N By doing so, the main NLMS filter transitions its tap to a noise state. Here, θ _TN is a regression to the noise level, and may be appropriately set in consideration of the noise level.

When the power of the k-th tap coefficient of the auxiliary NLMS filter in the transition state exceeds the predetermined threshold value θ _TS , the step size of the k-th tap of the main filter is set to μ _s which is the step size of the search state. In a multipath state, the signal fluctuates over time and the coefficients must be updated quickly. Therefore, the step size should be set to a large value so that the tap coefficient adapts quickly to changes in the channel.

Of course, in the present invention, the threshold value θ _TS , which becomes a judgment variable upon transition to the search state, prevents unnecessary repetition of the state transition due to the hysteresis effect in the relationship of θ _TS > θ θ _NT .

In addition, when the power (│W _{s , k} | ² ) of the auxiliary filter measured in the transition state is in the relationship of θ _TS <│W _{s , k} | ² <θ _TN , the corresponding tap of the main filter is a step in the transition state. It will keep the size.

When it is determined that the multipath does not exist in the search state, the signal returns to the noise state through a path opposite to the above-described transition to the search state. For example, when the power of the k-th tap coefficient of the auxiliary filter is smaller than the threshold θ _ST , the k-th tap coefficient of the main filter transitions to the transition state, and the step size is set to the value (μ _T ) of the transition state.

In the case where it is determined that the sensed multipath does not exist after a lapse of time, the transition from the search state to the transition state is the average value of the tap coefficient power of the auxiliary filter for a predetermined time (e.g., 1 slot) (∑│W _{s , k} | ² ) It is desirable to determine the transition or not based on this.

Because, due to the nature of time-varying radio fading, it is advantageous to improve the performance of the filter device so that the transition in the search state does not occur frequently because the filter device may return to the multipath environment in a short time. Of course, it will be appreciated by those skilled in the art to which the present invention pertains that the average value for a predetermined time may be used as the power value of the auxiliary NLMS filter, which is the basis for judgment upon transition to each state. will be.

Meanwhile, as described above, the return from the transition state to the noise state may be made by comparing the tap coefficient power and the transition threshold value θ _TN .

Through the above-described method, it is possible to detect the tap coefficient state of each tap of the auxiliary NLMS filter and control the step size of the tap of the corresponding main NLMS filter. Therefore, in the present invention, while the tap coefficient of some taps of the main NLMS filter is maintained in the search state, the remaining taps may be kept in a noise state. Therefore, the filter device of the present invention maintains the step size of the main NLMS tap corresponding to the multipath channel to a large value, thereby exhibiting a fast convergence characteristic in response to the fast time varying characteristics of the channel, and the remaining taps are kept at a low value as a whole in a steady state. Error characteristics can be improved.

Meanwhile, the present invention may optimize the step size by estimating the step size of the main filter in a state where multiple paths exist, that is, in a search state. For example, estimation of the Doppler frequency may be used to optimize this step size. That is, the Doppler frequency of each coefficient can be measured by measuring the change amount of the complex coefficient of each tap, and this value can be used as a measure of the change amount of the radio channel over time. Various methods may be used for estimating the Doppler frequency, for example, a known method of measuring a real zero / imaginary zero cross rate (N _ZCR ) of a complex coefficient represented by the following equation may be used.

f _{D , max} = N _ZCR * √2

Where N _ZCR = min {N _{ZCR , Re} , N _{ZCR , Im} }

By the Doppler frequency estimation, the convergence speed can be adjusted by increasing the step size of the tap when the Doppler frequency is large and decreasing the step size of the tap when the Doppler frequency is small. This means that the stem size of the search state is not fixed to a specific value μ _s but can be set to μ _s1 , μ _s2 , μ _s3 etc. according to the measured Doppler frequency.

 Although a method of individually controlling the step size of the filter in the multipath environment based on the interference elimination repeater has been described, the present invention can be used for radio channel estimation of not only an interference elimination repeater but also a general mobile communication base station / terminal. In addition, although the above-described embodiment illustrates an NLMS filter, it will be appreciated by those skilled in the art that the present invention may be applied to an LMS filter.

1 is a diagram schematically showing the configuration and operation of an NLMS filter device used in an interference repeater.

2 is a view schematically showing the configuration and operation of the NLMS filter apparatus according to the present invention.

3 is a diagram illustrating a tap coefficient state transition of an NLMS filter that is variably controlled according to the present invention.

<Short description of the symbols in the drawings>

10, 100: NLMS filter

120: Secondary NLMS Filter

200: feedback channel

Claims (14)

An LMS filter apparatus having a main LMS adaptive filter to perform radio channel estimation, And an auxiliary LMS filter for updating tap coefficients of the main LMS adaptive filter. The method of claim 1, Wherein the main LMS filter and the auxiliary LMS filter have the same number of taps. The method of claim 1, And a step size for each tap coefficient of the auxiliary LMS filter is fixed. The method of claim 1, And controlling the tap size of the corresponding main LMS filter on the basis of the state of the tap-by-tap coefficient of the auxiliary LMS filter. The method of claim 4, wherein And controlling the tap size of the corresponding main LMS filter according to the power of the tap-by-tap complex coefficient of the auxiliary LMS filter. The method of claim 1, And the main LMS filter and the auxiliary LMS filter are normalized LMS filters. Detecting a coefficient state of a plurality of taps of the filter with respect to a signal input to a first adaptive filter; and And varying the step size of the corresponding tap of the second adaptive filter based on the counted state of each tap of the first adaptive filter. Step size control method. The method of claim 7, wherein The coefficient state of each tap of the first adaptive filter is determined by the power of the complex coefficient of the tap of the first adaptive filter, step size control method of the adaptive filter device for wireless channel estimation. The method of claim 7, wherein The coefficient state of each tap of the first adaptive filter is determined by an average value of power of complex coefficients of the tap of the first adaptive filter measured for a predetermined time. Step size control method. The method according to claim 8 or 9, The step size variable step of the corresponding tab By increasing the step size of the tap of the second adaptive filter corresponding to the tap whose power or average power of the complex coefficient among the taps of the first adaptive filter exceeds a preset first threshold value. A step size control method of an adaptive filter device for estimating a wireless channel. The method according to claim 8 or 9, The step size variable step of the corresponding tab Reducing the step size of the tap of the second adaptive filter corresponding to the tap whose power or average power of the complex coefficient among the taps of the first adaptive filter is less than a second preset threshold. Step size control method of an adaptive filter device. The method of claim 7, wherein The step size control of the adaptive filter device for wireless channel estimation, characterized in that the step size variable step of the corresponding tap is performed through at least one transition state having a predetermined step size between the current step size and the final step size. Way. The method of claim 7, wherein And the step size in the step size variable step of the corresponding tap is varied by Doppler frequency estimation. The method of claim 7, wherein An initializing step of setting an initial step size of each tap of the first and second adaptive filters, The step size control method of the adaptive filter device for wireless channel estimation, characterized in that the step size of the first adaptive filter in the initialization step is larger than the step size of the second adaptive filter.

Images