Adaptive smart antenna processing method and apparatus

FIELD OF THE INVENTION The present invention relates to a wireless communication system, and in particular, to determining the right for adaptive smart antenna processing in a wireless communication receiver having an antenna element array and an adaptive smart antenna processing device. .

Technical background A wireless communication system including a communication station having an antenna array and an adaptive smart antenna processing device is well known. This communication station is sometimes called a smart antenna communication station. When receiving a signal from a subscriber unit, the signals received by each antenna array element are combined by an adaptive smart antenna processing device to provide an estimate of the signal received from a specific subscriber unit. Using smart antenna processing including linear space processing, each complex value (that is, including in-phase component I and quadrature component Q) signals received from the antenna element is weighted in amplitude and phase according to weighting factors, and then the weighted signals are summed to provide estimated value. In this way, the adaptive smart antenna processing device can be described by a set of complex-valued weights, each weight corresponding to an antenna element. Thus, a single complex-valued vector of m elements can be used to describe these complex-valued weights, where m is the number of antenna elements. This can be extended to include spatio-temporal processing, where the signal on each antenna element is not simply weighted in amplitude and phase, but filtered by some complex-valued filter, usually for time equalization. Each filter can be described by a complex-valued transformation function or a convolution function. In this way, the adaptive smart antenna processing of all the oscillators can be described by the complex-valued in-vectors of m complex-valued convolution functions.

Several methods are known for determining the weight vector of the received signal. These methods include methods to determine the direction of arrival of signals sent by subscriber units, and methods to utilize spatial characteristics of subscriber units, such as spatial signatures. For example, regarding the method of using the direction of arrival, see

00819675.3 See U.S. Patent Nos. 5515378 and 5642353 entitled "Space Division Multiple Access Wireless Communication System" granted to Roy et al.; for the method of using spatial signatures, refer to the "High-efficiency Spectrum High-Capacity Wireless Communication System" granted to Barratt et al. US Patent No. 5,592,490, and US Patent No. 5,828,658 entitled "High-efficiency spectrum and high-capacity wireless communication system with time-space processing" granted to Ottersten et al. The so-called "blind" method determines the right based on the signal itself, without resorting to training signals. That is, it is uncertain what weight can best estimate the known symbol sequence. This method usually uses some known characteristics of the signal sent by the subscriber unit to determine the best weight to use. The specific method is to force the estimated value to have this Characteristics, so called characteristic recovery methods. Characteristics recovery methods can be divided into two groups. Partial characteristic recovery methods recover one or more usually simple signal characteristics, instead of completely reconstructing the modulated reception, for example, by demodulating and then re-modulating Signal. The "decision-directed" (DD) method constructs an exact copy of the signal by making symbolic decisions (such as demodulation) on the received signal.

An example of the first group of partial recovery methods is the constant modulus (CM) method, which is suitable for communication systems that use modulation schemes with constant modulus. These modulation schemes include, for example, phase modulation (PM), frequency modulation (FM), and phase modulation (FM). Shift keying (PSK) and frequency shift keying (FSK). See, for example, JR Treichler; ML Larimore: New processing technology based on constant modulus algorithms (IEEE Transactions on Acoustics, Speech, and Signal Processing, ASSP-33, Issue 2, pages 420-431, April 1985). Other partial characteristic restoration techniques include techniques to restore the signal's spectral characteristics, such as spectral self-coherence. The spectral coherence recovery technique utilizes the known spectral coherence characteristics of any signal received at the antenna array. For example, in some cases, it can be assumed that the signal is cyclostationary, that is, it has a periodic autocorrelation function. Other methods include restoring higher-order statistics, such as moments or cumulants. See for example:

B. Agee, S. Schell, W. Gardner's spectral self-coherent recovery: A new method for blind adaptive signal extraction using antenna arrays" (Proceedings of the IEEE Vol. 78 No. 4, April 1990); and award US Patent No. 5,260,968 entitled "Method and Apparatus for Multiplexing Communication Signals by Blind Adaptive Spatial Filtering" by Gardner et al.; and entitled "Apparatus and Method for Extracting Self-coherent Recovery Signals" issued to Gardner et al.

00819675.3 Patent No. 5,255,210.

The decision guidance method uses the fact that the modulation scheme of the transmitted subscriber unit signal is known, and determines the right to generate a signal with the required modulation scheme ("reference signal"). If it is transmitted by a remote user, it will be Signals are generated on the antenna elements "close" to the actual received signal in the array, and the reference signal generation includes symbol decision making. For an introduction to the system using decision-making guidance power determination, see, for example, Barratt et al.s U.S. Patent Application 08/729390 (October 11, 1996) entitled "Method and Device for Decision-Guide Demodulation Using Antenna Array and Space Processing" Submitted); and Petrus et al.'s 09/153110 (September 15, 1998) entitled "Method for generating reference signals in the presence of frequency offset in a communication station using spatial processing"_β As we all know, some iterative methods, including partial recovery methods, such as CM methods, even for low signal-to-noise ratio (SNR), low signal-to-interference impulse-to-noise ratio (SINR), and communication systems where subscriber units move at high speeds. The high fading conditions encountered will also converge. This type of method is referred to herein as "the iterative weight determination method with good convergence characteristics". However, a method with good convergence characteristics can use many iterations to converge. For example, the CM method can use many iterations to converge, but it may not converge fast enough in an actual system. For example, in a highly mobile system, it is better to use the current burst weight vector derived from the current burst data. This means fast calculation of weights, which may not be possible with the CM method. On the other hand, the decision guidance method is a kind of rapid convergence under initial conditions, such as the initial signal-to-noise ratio (SNR) and the signal-to-interference impulse-to-noise ratio (SINR) are high, or the initial weight vector is close enough to the correct value. An instance of the method. The method of fast convergence when the initial weight vector is close enough to the correct value is referred to herein as the "fast convergence iteration weight determination method". Fast convergence methods, such as the DD method, are increasingly used in communication stations based on smart antennas. When this method fails in so-called low SINR or high fading conditions, the method may not converge. This problem becomes particularly serious in communication systems where many users face high co-channel interference, that is, when When receiving a signal from a specific subscriber unit, the interference of the signal in the regular channel from other subscriber units is very high. It includes several receiving communication stations, each communication station and its cell.

00819675.3 In the case of a cellular system where the first group of user units communicate, such other user units are from the same cell or neighboring cells.

In theory, adaptive smart antenna processing allows more than one communication link to exist in a single "regular" communication channel, provided that user units sharing the same regular channel can be spatially (or spatially and temporally) distinguished. Conventional channels include frequency channels in frequency division multiple access (FDMA) systems, time slots in time division multiple access (TDMA) systems (usually including FDMA, so to be more precise, conventional channels are time slots and frequency slots), and codes Code in Division Multiple Access (CDMA) system. It can be said that a regular channel is divided into one or more "spatial" channels. When there is more than one spatial channel in each regular channel, this multiplexing is called space division multiple access (SDMA). Here SDMA Used to refer to adaptive smart antenna processing including one or more spatial channels for each regular channel.

Fast convergence methods, such as decision guidance methods, will also fail when there is high co-channel interference in an SDMA system with more than one spatial channel per regular channel.

Therefore, there is a need in the art for an adaptive smart antenna processing method. For the SDMA system with one spatial channel for each regular channel and the SDMA system with multiple spatial channels for each regular channel, the difference between low signal-to-interference and noise In a relatively low environment or in a high-fading environment, the power of adaptive smart antenna processing is determined efficiently.

Therefore, there is a need in the art for a weight determination method that performs well under low SINR and high fading conditions and converges quickly (that is, through fewer iterations).

Therefore, there is a need in the art for a method that combines good convergence characteristics with fast convergence characteristics.

Therefore, there is a need in the art for a "blind" method (that is, a method that does not use training data) that combines good convergence characteristics (convergence at low SINR) and fast convergence.

SUMMARY OF THE INVENTION An object of the present invention is a weight determination method that combines the advantages of the method with good convergence characteristics and the advantages of the method with fast convergence.

Another purpose of the present invention is to perform well in the case of low SINR and high fading

00819675.3 The first and fast "blind" weight determination method and device that converge in a few iterations.

Another purpose is an adaptive smart antenna processing method and device. It has an SDMA system with a spatial channel for each conventional channel. It has high efficiency in an environment with a low signal-to-interference plus noise ratio or in a high-fading environment. To determine the right of adaptive smart antenna processing.

Another purpose is an adaptive smart antenna processing method and device. It has a SDMA system with multiple spatial channels for each conventional channel. In an environment with a low signal-to-interference plus noise ratio or a high fading environment, Efficiently determine the power of adaptive smart antenna processing.

Another object of the present invention is an adaptive smart antenna processing method and device, used to determine the adaptive smart antenna processing rights used in the current burst of data, and make these decisions based on the data in the current burst. The right is suitable for the current burst of data.

BRIEF DESCRIPTION OF THE DRAWINGS The present invention will be more fully understood from the detailed best embodiments of the present invention, but these embodiments should not be construed as limiting the present invention to any specific embodiment, but only for the purpose of description and better understanding . These embodiments are again illustrated with the aid of the following drawings: Figure 1 is a functional block diagram of a multi-antenna transceiver system, which may include a receiving right determination device according to various aspects of the present invention; Figure 2 is a transceiver including a signal processor A more detailed block diagram of the information machine, when a set of instructions is executed, the receiving right determination device according to various aspects of the present invention is implemented; FIG. 3 is a flowchart of an embodiment of the right determination method of the present invention; FIG. 4 shows the right determination method of the present invention The block diagrams of the tracking reference signal generator and demodulator used in the preferred embodiment; Figures 5A, 5B and 5C show the performance of the constant modulus method, the decision guidance method, and the hybrid method according to various aspects of the present invention, respectively; Figure 6 shows Multi-port right determination device and space according to the preferred embodiment of the present invention

00819675.3 A block diagram of the processor; Figure 7 is a flowchart of a preferred embodiment of the multi-user rights determination method of the present invention; Figure 8 illustrates the effect of timing offset on the performance of the CM method; and Figure 9 illustrates one aspect of implementing the present invention Block diagram of the device.

Detailed description of the best embodiment The method and device of the best embodiment of the base station architecture are implemented in a communication receiver, specifically, in a PHS-based antenna array communication station (transceiver), As shown in Figure 1, There are m antenna elements in the antenna array. In this particular embodiment, m=4. Although systems similar to those shown in FIG. 1 may be prior art, systems such as the system of FIG. 1 having elements that implement various aspects of the present invention through programming or hard wiring are not prior art. Moreover, the present invention is by no means limited to adopting PHS air interface or TDMA system, but any communication receiver including an adaptive smart antenna processing device. In FIG. 1, a transmitting/receiving ("TR") switch 107 is connected to the m-antenna array 103 and the transmitting electronic device 113 (including one or more transmitting signal processors 119 and m transmitters 120) and the receiving electronic device 121 (including m receivers 122 and one or more received signal processors 123), in order to selectively connect one or more units of the antenna array 103 to the transmitting electronic device 113 when in the transmitting mode, When in the receiving mode, one or more elements of the antenna array 103 are selectively connected to the receiving electronic device 121. Two possible implementations of the switch 107 are a frequency duplexer in a frequency division duplex (FDD) system and a time switch in a time division duplex (TDD) system. The preferred embodiment of the PHS of the present invention uses TDD. The transmitter 120 and the receiver 122 may be implemented using analog electronic devices, digital electronic devices, or a combination of the two. The receiver 122 of the preferred embodiment generates the feed to a One or more signal processors 123 digitized signals. The signal processors 119 and 123 can be static (always the same), dynamic (change according to the desired directionality) or intelligent (only change according to the received signal), and are adaptive in the preferred embodiment . The signal processors 119 and 123 can be the same or multiple DSPs that are programmed differently for reception and transmission.

00819675.3 The first device, or a different DSP device, or a different device for some functions, and the same device for other functions.

It should be pointed out that although Figure 1 shows a transceiver in which the same antenna element is used for receiving and transmitting, it is obviously also possible to use independent antennas for receiving and transmitting. Only receiving, only transmitting, or both receiving and transmitting can include adaptive smart antennas. deal with.

For example, the version 2 of the main standard of the Association of Radio Industries and Business (ARIB, Japan), the personal hand-held telephone system (PHS) described in RCR STD-28, is based on the technical standard of the PHS Memorandum of Understanding group (PHS MoU-see http:/ The variants described in /www.phsmou.or.jp) are all 8-slot time division multiple access (TDMA) systems with actual time division duplex (TDD). Therefore, the 8 time slots are divided into 4 transmission (TX) time slots and 4 reception (RX) time slots. This means that for any specific channel, the receiving frequency and the transmitting frequency are the same. This also means reciprocity, that is, assuming that the movement of the subscriber unit between the receive slot and the transmit slot is minimal, the downlink (from the base station to the user's remote terminal) and the uplink (from the user's remote terminal to the The propagation path used by the base station is exactly the same. The frequency band of the PHS system used in the preferred embodiment is 1895-1918.1 MHz. Each of the 8 time slots is 625 microseconds long. The PHS system has dedicated frequencies and time slots for the control channel on which call initialization is performed. Once the link is established, the call is transferred to the traffic channel used for regular communication. Communication takes place in any channel at a rate of 32 kilobits per second (kbps) called full rate. Communication below the full rate is also feasible. For those skilled in the art, It should be clear how to modify the embodiments described here to incorporate sub-full rate communications.

In the PHS adopted in the preferred embodiment, "burst" is defined as an RF signal of limited duration transmitted or received over the air during a single time slot. "Group" is defined as a group of 4 TX time slots and 4 RX time slots. A group always starts from the first TX time slot, and its duration is 8 X 0.625 = 5 microseconds.

The PHS system uses "/4 differential four-phase (or quadrature) phase shift keying (/4 DQPSK) modulation of the baseband signal. The baud rate is 192 kilobaud. That is, there are 19200 (? symbols per second.

Figure 2 shows a more detailed but still simplified block diagram of a PHS base station, which includes adaptive smart antenna processing, and on which the embodiments of the present invention are implemented. Furthermore, although

00819675.3 The first system with a structure similar to that shown in Figure 2 may be prior art, but systems such as the Figure 2 system with elements programmed or hard-wired to implement aspects of the present invention are not prior art. In Figure 2, multiple m antennas 103 are used, where m = 4₀More or fewer antenna elements can also be used. The output of the antenna is connected to the duplexer switch 107, which is a timing switch in this TDD system. When receiving, the antenna output is connected to the receiver 205 through the switch 107, and the RF receiver module 205 mixes down from the carrier frequency (about 1.9 GHz) to the intermediate frequency ("IF") in an analog manner. This signal is then digitized (sampled) by an analog-to-digital converter ("ADC") 209. The digital down-converter 213 then performs digital down-conversion to generate a four-times over-sampled complex value (in-phase I and quadrature Q) sampled signal. In this way, the elements 205, 209, and 213 correspond to the receiver 122 of FIG. For each of the m receive time slots, m down-converted outputs from m antennas are fed to a digital signal processor (DSP) device 217 (hereinafter referred to as "slot processor") for further processing. In the preferred embodiment, a commercial DSP device is used as the time slot processor, and there is one DSP for each received time slot.

The time slot processor 217 performs various functions including the following: received signal power monitoring, frequency offset estimation/correction, and time offset estimation/correction, including using the method according to an aspect of the present invention to determine the weight of each antenna element. Smart antenna processing of signals from specific remote users and demodulation of determined signals.

The output of the time slot processor 217 is a demodulated data burst corresponding to each of the m (=4) reception time slots. This data is sent to the main DSP processor 231 whose main function is to control all elements of the system and cooperate with higher-level processing, which is to process all the different control and service communication channels defined in the PHS communication protocol. signal. In the preferred embodiment, the main DSP 231 is also a commercial DSP device. In addition, the time slot processor sends the determined receiving right to the main DSP 231:

The RF controller 233 is connected to the RF system shown in block 245, and also generates some timing signals used by both the RF system and the modem. The RF controller 233 receives its timing parameters and other settings for each burst from the main DSP 231.

The transmission controller/modulator 237 receives transmission data from the main DSP 231. Send control

00819675.3 The first device uses this data to generate an analog IF output to be sent to the RF transmit (TX) module 245. The specific operations performed by the transmission controller/modulator 237 include converting data bits into complex-valued /4 DQPSK modulated signals, up-converting to IF frequency, weighting according to the complex-valued transmission weight obtained by the main DSP 231, and using digital-to-analog conversion A converter ("DAC") converts the signal into an analog transmission waveform to be sent to the transmission module 245. The transmitting module 245 up-converts the signals to the transmitting frequency and amplifies these signals. The amplified transmission signal output is coupled to m antennas 103 through a duplexer/timing switch 107.

The symbols are as follows. Assume that there are m antenna elements (m=4 in the preferred embodiment) and Z](/), Ζ2(ί)> are respectively after down-conversion, that is, in the baseband and sampling (in the preferred embodiment, it is four times over). After sampling), the complex-valued responses of the first, second, ... Π1-th antenna elements (that is, with in-phase I and quadrature Q components). In the above symbols, but not essential to the present invention, / is discrete. These m time samples can be represented by a single m-vector z(r), where z(/) in the z-th row is ζ©). For each burst, a finite number, for example, N samples are collected, so that zQ, scoop, ..%(/) can each be represented as an N-row vector and z(/) can be represented by an mxN matrix Z. In the following detailed description, the details of combining a limited number of samples are omitted, and it should be clear to those skilled in the art how to include these details.

Suppose that multiple signals are sent to the base station from several, say, a remote user. Specifically, suppose the involved user unit transmits the signal s(6 Adaptive smart antenna processing includes the received signal Z](/), Z2(r)...z_m(The I value and Q value of 0 are combined in a specific way to extract the estimated value of the transmitted signal s(/). These weights can be represented by the receiving weight vector corresponding to a specific subscriber unit, which is represented as the i-th element w" The complex-valued weight vector is called. The estimated value of the transmitted signal is: m

Extinguish) = £w; z,(/) = w: z(/) (1) ί=1 where grumble is the complex key, but the Hermitian transpose of the receiving weight vector W (ie, transpose He Fu Gong Ting). In an embodiment that includes spatio-temporal processing, each of the received weight vectors

00819675.3 The first element is a function of time, so the weight vector can be expressed as w"(/) with the Zth element w"(/). Then the estimated value of the signal can be expressed as:

Disaster) = £w"(/)*zQ /=1 where the operator "*" is a convolution operation. Spatio-temporal processing, such as the combination of time equalization and spatial processing, is particularly useful for wideband signals. The use of spatio-temporal processing to form signal estimation The value can be equivalently carried out in the frequency domain (Fourier transform). § (/), Z: (/) and w (/) in the frequency domain are respectively handsome), Z, and Qi), where Force is a discrete frequency value, m

Such as Σ%(£)Ζ©) (3)

Ϊ=1 Using spatio-temporal processing, the convolution operation of formula (2) is usually limited and performed on sampled data, which is equivalent to using a time domain equalizer with a limited number of equalizer taps to combine spatial processing and time equalization. That is, each guard has a limited number of /values, which is equivalent to having a limited number of force values for each% in the frequency domain. If the length of the convolution function called is «, the complex-valued in-weight vector W" is not determined, but the complex-valued m xn matrix is determined. Its columns are n values called ").

In the rest of the description, as long as the complex-valued receiving weight vector gas or its elements is mentioned, it should be understood that this can be used for spatial processing or a broad combination of the above-mentioned temporal and spatial processing to determine the weight matrix billion. Therefore, both spatial processing and spatio-temporal processing are referred to herein as adaptive smart antenna processing.

Determining the weight of the space The "blind" method of determining the weight for adaptive smart antenna processing is to not need to reconstruct the training data. The method of the present invention, similar to most blind methods, utilizes some knowledge of the format of the originating signal and limits the output signal to have one or more known input signal characteristics. The characteristics can be amplitude characteristics or some statistical characteristics, such as straight or cyclostationarity, correct modulation scheme, or reconstruction of an accurate copy. Such methods are sometimes referred to as feature recovery.

Methods with good convergence characteristics Methods with good convergence characteristics include partial characteristic restoration methods: by determining the ratio

00819675.3 The special stream and reconstructed signal reproduce one or more characteristics without attempting to reproduce some methods of exact reproduction. Such methods include methods of preserving the signal's amplitude (mode), directness, and spectral coherence (such as cyclic stability).

The constant modulus (CM) method is a very simple and effective technique, which is suitable for signals modulated by a scheme that obtains a constant amplitude signal. These include all forms of phase and frequency modulation, including the differential phase shift keying modulation of the PHS system used in the preferred embodiment. As described below, the CM method is also suitable for non-constant modulus signals. The CM method determines the weight that satisfies the following conditions: a) the constant amplitude (constant modulus) characteristic of the restored signal, and b) the generated signal, if it is transmitted by a remote user, should be "close" in the array to the antenna that actually receives the signal A signal is generated at the vibrator. Due to interference, fading, and timing offsets in those modulation schemes, amplitude variations may be introduced. In the modulation scheme, the constant modulus characteristic depends on accurate timing offset correction, including, for example, that the constant modulus characteristic is only kept at the maximum in the wave characteristics. DQPSK modulation scheme of the preferred embodiment. In the presence of co-channel interference, the CM method often picks up the strongest signal, regardless of whether it is the desired signal or co-channel interference. Even if the strength of the desired signal is only 0.5dB greater than the strength of any interfering signal, the CM method will still correctly pick up the strongest, that is, the desired signal. That is, the CM method has very good convergence characteristics.

There are many variations of the CM method. They usually minimize the cost function of the general form: = (4) where Nine.) represents the statistical expected value operation, P and α are positive integers, generally 1 or 2. Obviously for those skilled in the art, in practice, this statistical operation is replaced by a form of averaging or accumulating samples (for example, by comparing the sub-values of all samples in the burst in the best embodiment). The sum of the sample set of the set). Furthermore, it is obvious that more items can be added to the cost function of formula (4), which does not go beyond the scope of the present invention. For example, you can add an item to limit the size of the weight vector. For an example of a cost function (non-CM cost function) with such additions, please refer to the above-cited U.S. Patent Application 08/729390°. The signal interval (1) is the normalized copy signal used in the cost function (referred to as the "reference signal). "). That is, for

00819675.3 The reference signal of the first determination weight is the weighted sum of the received antenna signals that are subsequently normalized. The determination of weights is to determine the set of weights that minimizes the cost function in formula (4).

The CM method is also suitable for non-constant modulus signals. For example, see J. Lundell and B. Widrow: Application of constant modulus adaptive beamforming devices on constant and non-constant modulus signals [Proceedings, 1988 Asilomar Conference on Signals, Systems and Computers (ACSSC-1988), No. 432 -436 pages, 1988]. Lundell and Widrow use a cost function, such as formula (4) when p=q=2 (this is called the 2-2 CM method), which shows that any constant and non-constant modulus signal can be recovered by this 2-2 CM method. As long as the ratio of the fourth moment to the square of the second moment (this ratio is called kurtosis) is less than 2. For example, it is well known that an M-quadrature amplitude modulation signal (M-QAM) has a kurtosis of about 1.4-L2/(M-1), so the kurtosis of any QAM signal is always less than 1.4. So the CM method is suitable for this kind of signal.

The specific method with good convergence characteristics, at least one iteration of the CM method is used in the best embodiment, and the realization of the CM method uses the value of ρ as \ and q as 2 in the formula (4). When the present invention is applied to a non-constant modulus signal and the CM method is adopted, other values of ρ and q may also be used, for example, p=q=2. The best implementation can also be a block-based approach. That is, the block of the signal received by the antenna is weighted, and the data block is used to determine the weight. A block is a subset of the samples in the burst. Specifically, it is best to use 75 samples of 120 PHS burst symbols, where these 75 symbols are in the payload in the middle of the PHS burst. Utilizing the data in the payload advantageously ensures that the data used for calculating the rights of any remote user is different from the data used by another user unit. There are up to 88 such payload samples in a PHS burst.

When p = l and q = 2, this method is called the least squares constant modulus method, including the following steps:

1. Initialize the weight vector. For example, use>v_r__inittal = [10 0...0J, where x'represents the transpose of X. In an improved embodiment, R" = ZZ corresponding to the largest singular value of Z is used^HThe largest eigenvector of. In another embodiment, the weight vector derived from the previous burst is used;

2. For the samples involved, perform copy signal and normalization:

00819675.3 p.

<img file="CN100372404C_D0001.tif" />

(5)

3. Use the least square method to calculate the weight vector. which is,

N w =argmin Y(s')-w; z(/))2 (6)

W, ί=1 where N is the number of samples used in the calculation. The solution of formula (6) is:

=R; h

N where Rzz = ZZH, G=, z(/)s'); and N is the number of samples used; and f=1

4. Repeat steps 2 and 3 until convergence is reached.

It should be pointed out that in the calculation of step 3, in fact, the overall conversion factor is not important. It is best to apply the conversion factor of ownership to the system as a gain in combination.

It should be noted that the CM method can be extended to spatio-temporal processing. A well-known method uses the 2-2 CM method, which shows that under certain assumptions that usually meet the actual situation, the CM method used for space-time weight determination (that is, for weight matrix determination) must converge. See CB Papadias and A. Paulraj's space-time constant modulus algorithm for SDMA systems (Proceedings, LEEE 46th Vehicular Technology Conference, pp. 86-90, 1996), but this method is not based on block data. However, it is possible to reformulate the problem based on different sizes of matrices and vectors, so as to easily modify the spatial weight determination method for the temporal and spatial processing of the weight matrix. For the entire description, it is assumed that m is the number of antenna elements and N is the number of samples. Let η be the number of time equalizer taps of each antenna element. It is possible to rewrite the vector of N samples in each row of (plus X received signal matrix Z) into an offset version of the first row of n rows to generate a received signal matrix Z of size (mn χ Ν). When multiplied by the Hermitian transpose of a weight vector of size (mn X 1), an estimated received signal row vector of N samples is generated. Therefore, the space-time problem has been re-formulated as a weight vector determination problem. For the CM method, the formula In (7), for example, the weight vector is a long "weight vector with size (afterload X 1), a matrix with size (mn X mn)^, and 1 is a long vector with size (add X 1)). Re The permutation item provides the required (edge χη) weight matrix.

Because in the preferred embodiment, in order to maintain the CM characteristics, the sampled data needs to be large

00819675.3 On-baud, when performing step 2, timing offset estimation and correction are performed. In this case, sampling and interpolation in time can be included, since each antenna from the antenna array 103 The samples of the signal received by the vibrator are oversampled and may include some timing offset. Therefore, the variable/in formula (5) and formula (7) represents the approximate time in baud of the sample value. Obviously and as described in the above-cited U.S. Patent Application 09/153110, the timing offset estimation/correction (may include sampling/interpolation) can be performed on m signals before the signal copy operation, or after the signal copy operation carried out.

Simulations have been performed to determine the accuracy level of timing offset/wave characteristic estimation. Figure 8 shows the results. In this simulation, when the signal is not just sampled on the wave characteristics, but is offset by some timing offset from the ideal wave characteristics, the output SINR obtained by using the CM weight is calculated, where the timing offset is based on the baud The step size of 1/8 varies from -1/2 to +1/2 baud. The results show that even for a signal offset of ±1/8 baud, the output SINR is only reduced by 0.6 dB_oAlthough this figure is for the test example shown, the conclusion is that the accuracy of the timing offset correction of the CM method does not need to be very high. Therefore, a simple method can be used to perform offset correction/sampling/interpolation in order to generate samples approximately aligned with the baud.

It should also be pointed out that the timing offset correction (including sampling/interpolation) of the oversampled signal may not be necessary for all modulation schemes with constant modulus characteristics. For example, the "European Digital Cordless Telecommunications" (DECT) standard and the "Global System for Mobile Communications" (GSM) standard use Gaussian Minimum Frequency Shift Keying (GMSK) signals that always have a constant modulus, so for the CM method in those examples , Timing offset correction is not needed.

The realization of the least square CM weight determination method is quite simple. Because demodulation is not performed, frequency offset estimation and correction and demodulation are not required. In an embodiment of the present invention, although it is not required, when the CM method is implemented, frequency offset correction is still performed. An additional feature of simple characteristic restoration methods such as the constant modulus method is that convergence occurs even if the value of the signal-to-interference plus noise ratio (SINR) is very low.

The main disadvantage of methods with good convergence characteristics, such as simple characteristic restoration methods, is that they may take many iterations to converge. In a normal system, the processing power, such as the processing power of the DSP, is extremely limited. Therefore, for example, in order to use the weight vector in the current burst

00819675.3 No. The CM method may not be convergent within a certain amount of time.

The fast convergence method is different from the partial feature restoration method. The fast convergence method, such as the decision-guided method, converges very quickly. Using the decision-guided method, the characteristic of recovery is a complete copy of the original signal with the correct modulation scheme. That is, the signal copy operation, as in formula (1), estimates the received signal, demodulates the signal, and constructs a reference signal with the correct bit stream. In order to proceed smoothly, any frequency and timing offsets need to be corrected when constructing the reference signal. The correct weight generates a reference signal that approximates the transmitted signal. The solution can include one or more iterations to obtain the "best" weight. Although the timing offset correction (including any sampling) and the frequency offset correction are shown below as occurring after the signal copy operation, it is obvious that one or more of them can be performed before the signal copy operation. For examples of these operations before and after signal replication and a detailed description of the decision-making guidance method, see the co-owned U.S. Patent Applications 08/729390 and 09/153110 cited above. When the least squares criterion is adopted, the usual method includes the following steps:

1. Initialize the weight vector. For example, using W_rinitial = [ιοο...ο], where X'represents the transpose of X. In an improved embodiment, the R corresponding to the largest singular value is used_zz =ZZ^HSingular vector. In yet another embodiment, the weight vector derived from the previous burst is used. As described below, one aspect of the present invention includes adopting a decision guidance method after adopting a partial characteristic method. In this case, when implemented in any of the embodiments of the present invention, the final weight vector is used (that is, the partial characteristic restoration method is used);

2. Perform signal copying s(t) = w^z(t) (8) If the sample is initially oversampled, then sampling/interpolation is performed subsequently (in another scheme, sampling/interpolation can be used in the copy signal Before calculation);

3. Estimate timing and frequency offset to generate a signal with correct timing and frequency offset;

4. By making a symbol decision (ie, demodulation), the interval has the correct bit stream and the same modulation scheme, and has the same signal as the signal transmitted from a specific user to the receiver.

00819675.3 The timing and frequency offset to determine the reference signal S seed (/);

5. The weight vector is calculated by the least square minimization of the excess. which is,

N ² Called =arg min £ \s_ref ¢)-protect z(/)| (9) Its solution is, called = (1°) where R_zz = ZZ^HWabata=£z(/)¢); and /=1

6. Repeat steps 2, 3, 4, and 5 until convergence is reached.

It should be noted that steps 2, 3, and 4 need to correct the signal with respect to frequency and timing offset, so that the correct demodulation decision can be made in step 4. At the same time, step 5 usually needs to re-introduce the correct frequency and timing offset to make the cost function The reference signal and the copied signal have the same timing and frequency offset. It should also be pointed out that the minimization of formula (9) can also include other items, such as the weighted norm term of the weight vector, in order to set constraints on the norm of the weight vector, as described above for the CM example and the US application 08 cited above. The discussion in /729390. See also the co-owned US patent applications 08/729390 and 09/153110 cited above for a detailed description of how to determine the reference signal (step (4)). This will be explained below with reference to FIG. 4.

It should be pointed out that the decision guidance method can be easily extended to determine the weight matrix of spatio-temporal processing, for example, rearranging some items for the CM method as described above, and other methods well known to those skilled in the art. Therefore, the present invention also includes a method for determining a weight vector and a weight matrix for spatio-temporal processing.

Therefore, the decision-guided method reproduces an exact copy of the signal assumed to be transmitted to the receiver, while the partial characteristic recovery method reproduces one or more simple characteristics, such as the correct amplitude. The decision-guided system works very well and converges after very few iterations in a fairly high SINR environment. However, these methods are sensitive to initial conditions and cannot even converge when the initial SINR is very low. This is common in highly mobile cellular systems and other systems that exhibit fading.

It should be pointed out that the iteration of the CM method is usually computationally faster than the iteration of the decision-guided method

00819675.3 The cost is low because no frequency offset correction or demodulation is required.

Best method: Single user One aspect of the present invention is a weight determination method, which includes M iterations of an iterative weight determination method with good convergence characteristics (such as a partial characteristic restoration method, preferably a CM method), and fast The second number ν of a combination of convergence methods and subsequent implementation of fast convergence methods (such as decision-guided methods)₂In order to obtain the advantages of good convergence characteristics and fast convergence. N\ CM iterations make the N of the decision-guided method₂The starting condition of each iteration is in the most likely range of the rapid convergence of the decision-guided method. The best embodiment uses one iteration of the decision-making guidance method (M =1), while the other implementation scheme uses two iterations (he=2). This method can be re-described as performing iterations of the iterative weight determination method with good convergence characteristics until the switching criterion is met, and then, starting from the weights obtained by the method with good convergence characteristics, perform several iterations of the fast convergence method. In one example, the switching criterion is a clearly defined number of iterations. In another preferred embodiment, N] is not explicitly specified. On the contrary, the switching criterion is the SINR threshold of the copied signal, and when the SINR estimate is equal to or exceeds the threshold, the switching to the decision-guided method is performed. In this way, a sufficient number of M CM iterations are used in order to obtain enough to ensure that only N₂A further iteration is the SINR at which the decision-guided method converges.

There are many ways to determine the SINR estimate. In the preferred embodiment, the method used is as described in Yun's US Patent Application 09/020049 entitled "Power Control Using Signal Quality Estimation for Smart Antenna Communication Systems" (filed on February 6, 1998). The realization of the signal quality estimation method will now be explained.

Let N represent the number of burst samples to be used for estimation. First, the sampling modulus information is extracted by summing the squares of the in-phase and quadrature signals (the real and imaginary parts of the signal s(r)). Then the average power and average square power are determined by the average value of the number of samples that are expected to be calculated.

--- Ί N Nao EryouΣ?Ρ)+ί2Ρ) (11)

Ν t=l ___, Ν2 ^=νΣ(^/2ω+β²ω) (12)

00819675.3 First, it should be pointed out that once the instantaneous power (/) = /2(/) + 02(/) is determined, the square power (/)=[(/)]2 only needs to be performed for each sample. Additional multiplication operations, and to determine the estimated signal to interference plus noise ratio, it is best to use at most one square root operation, using

<img file="CN100372404C_D0002.tif" />

The ratio %* and the quantity A are sometimes called kurtosis. This preferred embodiment of signal quality estimation is insensitive to frequency offset, so it is a particularly attractive method for use with the CM method, which is also insensitive to frequency offset.

In another implementation of the present invention, other methods for determining the quality of the post-copy operation signal can also be used.

The method of determining the right of a single user is illustrated in the flowchart of FIG. 3. At 303, an initial weight vector is formed. It can be [10 0...0]', or in an improved embodiment, the R corresponding to the largest singular value is used_zz =ZZ^HSingular vector. In yet another embodiment, the weight vector from the previous burst is used. The copy operation 305 is now performed according to formula (1), but in the preferred embodiment, only the middle part of the burst is used, preferably only 75 symbols (300 samples) in the payload part of the middle of the burst are used. At 307, the timing offset is used to correct the output. Any timing offset correction method can be used. As discussed above, the timing offset correction does not need to be very accurate. The best method is as described in the above-cited U.S. Patent Application 09/153110. The copy operation and timing offset correction operation can be combined. Any necessary sampling and interpolation are inherent in the timing offset correction calculation, but it is not explicitly shown in Figure 3, so after step 307, the data includes approximately 75 of the wave characteristics from the 75 symbols in the middle of the current burst. Complex value (I and Q) samples. At 309, the kurtosis described above is preferably used to estimate the SINR of the replicated signal. In step 311, it is determined whether the SINR exceeds the threshold

00819675.3 The first value SNR. If not, in step 313, the iteration of the constant modulus method is performed using the least square cost function criterion described in the above formula (6) and formula (7). The method then returns to the copy operation of step 305 to perform another iteration. On the other hand, if it is determined in step 311 that the SINR threshold has been exceeded, in steps 315 and 317, other iterations of the decision guidance method are executed, including the frequency offset correction in 315. Any frequency offset correction method can be used, and the best method is the method described in the above-cited U.S. Patent Application 09/153110. Similarly, for decision-guided adaptation, including generation of reference signals, any method can be used, and the preferred embodiment adopts the method described in the above-cited US Patent Application 09/153110. When the weight is determined, in the preferred embodiment, only a subset of samples in each burst is used. Therefore, in the copy operation and demodulation step 318, the final weight vector is now used for the entire burst. In this embodiment, step 318 includes timing and frequency offset determination and correction, and demodulation, and the architecture described in FIG. 4 is preferably used. The output of the decision-guided adaptation is signal 319.

As part of the decision-guided adaptation in step 317 (using a part of the burst data) and the best embodiment for generating the reference signal for demodulating all burst data in step 318, it is best to use a reference signal. Generating architecture and methods, which include a tracking mechanism (preferably sample-to-sample), by relaxing the phase of the signal ideally ahead of the previous reference signal sample to the phase of the replicated signal at the same value point, forming The phase of the reference signal at the sample point, and the replica signal is formed based on the received antenna signal. The reference signal is constructed on each sample point by the following steps: construct a normal signal sample value at the same sample point according to the replicated signal, and the ideal signal sample has a value determined based on the replicated signal at the same point The phase of the ideal signal sample at the initial symbol point is set as the initial ideal signal phase, and then the phase of the ideal signal sample is relaxed to the phase of the replicated signal sample to obtain the phase of the reference signal. The phase of the ideal signal is determined based on the phase of the reference signal at the same value point for which the phase is determined and the decision based on the replicated signal. In one implementation, the reference signal is determined in the forward time direction, and in another implementation, the reference signal samples are determined in the reverse time direction. In one solution, the step of relaxing the phase of the ideal signal sample to the phase of the replicated signal corresponds to

00819675.3 is the filtered version that adds the difference between the phase of the reproduced signal and the phase of the ideal signal. In another scheme, the step of relaxing the phase of the ideal signal sample to the phase of the replicated signal corresponds to forming a reference signal sample by adding a filtered version of the difference between the replicated signal and the ideal signal to the ideal signal sample.

This will now be described in detail with reference to Fig. 4 taking the /4 DQPSK PHS signal as an example. It is obvious to those skilled in the art that other modulation schemes can be obtained by modification. The phase detector unit 403 detects the phase difference 405 between the replica signal 401 (corrected for timing and frequency offset) and the previous reference signal 417. The phase difference signal 405 is fed to the limiter 407 to generate the decision phase difference 419₀ /4 The correct phase difference of DQPSK is (2/-1)/4, Z=1, 2, 3, or 4, and it is the phase difference between the previous reference signal sample and the ideal signal. In block 409, it is subtracted from the actual phase difference 405 to generate an error signal 411. This error signal is filtered by the filter 413 to generate a filtered error signal 415. It is a filtered error signal used to adjust the phase difference 419 to be closer to the actual phase difference 405. The corrected phase difference 421 is then used in the frequency synthesizer/phase accumulator 423 to generate the reference signal 429. It is the previous sample value 417 of the reference signal 429 used by the phase detector 403, so these signals represent a unit time delay 425 between them. The sign of signal 430 (signal 319 after one iteration) is determined by block 427. Mathematically, if the taste (/) represents the complex sample value of the reference signal on the wave characteristic/ and represents the phase, then the input b of the phase accumulator 423_R (/)-% (/ -1) is: filter{ d_tdeal (n) -decide {d such as (»)}}+decide{ d_ideal (n)}, where decide{%»} is the output of limiter 407, and for /4 DQPSK, it is equal to (2/-1)/4, i=1.2, 3, or 4. Here, the "ideal" complex value sample point plus "(/) is defined as: also (0)yi(0)=b(0), where b(/) is the sample value of the input signal 401, and 4 is (Ο is the phase difference between the current input sample value and the previous reference signal sample value: α such as (r)M (11)=of)shen(/-1)], where * represents the complex norm." The "ideal" signal is a reference for the phase advance by one rational amount

00819675.3 No. signal, the amount depends on the decision made according to α such as ¢). In other words, square idea/C)=square R(;-l) + -1)/4, i=l, 2, 3, or 4.

To obtain this reference signal, now relax the phase of scoop (/) to the phase of b(/), specifically by measuring the phase error between force and scoop . Filter, and add the filtered amount to the phase of square (/). Another example is the amount (blood)-b_ideal ¢)) instead of the phase difference for filtering. The filter is preferably a constant of proportionality. High-order filters can also be used. Mathematically, in one embodiment, bQ = 2 +filter{b(/)Yi (/)}, while in another embodiment, the architecture of Figure 4 can be slightly modified to adopt bQ = ^_a;(0+filter{ 0(/)-^(/)} In the preferred embodiment, the method of the flowchart of FIG. 3 includes a tracking reference signal generator, which is a time slot used as a signal processor (DSP) device The processor 217 is implemented in the form of an instruction set.

For the system, the simulation of the method in Fig. 3 is performed, but the initial weight vector, R_zz =ZZ^HThe eigenvector of corresponds to the maximum eigenvalue. The simulation is performed for the PHS base station of Fig. 2 with four antenna elements. The input signal of each antenna has a signal-to-noise ratio (SNR) of 11.9dB. The input carrier-to-interference ratio (CIR) is 1.ldB, which corresponds to the initial replica signal SINR of 0.8dB. Although the conventional PHS burst has 120 symbols, only the central 75 symbols are used for weight calculation. All calculations are performed offline using MATLAB environment (Mathworks Inc., Natick, MA). Figures 5A, 5B, and 5C show the results, which compare the decision-guided method, the Nie Xiao-square CM method, and the combined method of the present invention under low SINR conditions (in this case, after the first copy operation using the initial weight vector After copying, the SINR is about 0.8 dB) convergence characteristics. Use the SINR estimation method to measure and plot the output SINR (dB) after each iteration. The first SINR value displayed is when SINK is estimated after the copy operation using the initial weight vector (the first singular vector of &amp;z). This is the same for all three methods. As shown in Figure 5A, even after 10 iterations, the decision guidance method does not converge. Figure 5B shows that the CM method converges very slowly, and the output (estimated) SINR continuously increases with iteration. The best SINR is 18 dB,

00819675.3 First, the CM method requires more than 10 iterations to converge to this optimal value. Figure 5C shows that the method of the present invention works with a switching output SINR threshold of 7.5 dB. It should be noted that this result is completely consistent with that of FIG. 5B at first, but diverges after switching to the decision-making guidance method (after switching, the result of FIG. 5B is represented by a dotted line). After the decision-guided method starts, the method reaches the optimal SINR after only 2 iterations of the decision-guided method, and even after only one iteration of the decision-guided method, it is very close to the optimal value. In a word, the method of the flowchart in Fig. 3 converges within 5 iterations, while it takes more than 10 iterations if only the CM method is used.

Multi-port architecture. When there are multiple, for example, M user units in the same regular channel (ie co-channel users) within and outside the cell, the preferred embodiment of the present invention adopts a multi-port architecture, each " The "ports" respectively form duplicate signals and track a single unit in the M subscriber unit, so the subscriber unit of any port becomes the co-channel interference of the remaining M subscriber units and their corresponding ports. Therefore, only co-channel users with signal components above a certain noise floor received at the antenna element are tracked. The number of such users can be estimated. Given any burst (matrix 2), you can check R_zz =ZZ^HEigenvalues and can perform first-order estimation. Any order estimation method can also be used. For example, the Rissanen minimum description length (MDL) standard or the Akaike information theory standard are well known. For an overview of techniques for determining the number of effective co-channel users, see section 3.8 of Rias Muhamed and TS Rappaport, "Estimation of Direction of Arrival Using Antenna Arrays" [Technical Report MPRGTR-96-03, Mobile and Portable Radio Research Group, Bradley Department of Electrical engineering, Virginia Polytechnic Institute, January 1996], and Rias Muhameds master's thesis using antenna array direction of arrival estimation (Bradley Department of Electrical Engineering, Virginia Polytechnic Institute and State University, Blacksburg, VA 24061, USA ). The preferred embodiment uses the minimum description length standard.

Although the preferred embodiment involves estimating and then tracking all "effective" co-channel users, in another embodiment, a "good" wireless design environment is assumed,

00819675.3 First, assume that the co-channel users who are not communicating with the same base station are far away, so the only valid co-channel users are those users who share the same regular channel and communicate with the base station. That is, the subscriber units of different spatial channels in the regular channel. In this case, the dish is known.

For example, consider when there are two known (by estimation or by existing information) co-channel users in this environment (ie, heart = 2). When tracking one of these two subscriber units, the other subscriber unit is the interfering party. Therefore, in this architecture, the required signals and effective interference for communication in the same regular communication channel are tracked at the same time. For example, in a fading environment encountered when a subscriber unit moves quickly, the carrier-to-interference ratio (CIR) may be very low, and the instantaneous CIR may fluctuate over a wide range. Therefore, at any point in time, in any given burst, the desired signal may be weaker than any of its interference signals, and the port of the desired signal may be locked by the interference. That is, it may start to track the interfering party instead of the desired remote user.

Fig. 6 shows a block diagram of the multi-port adaptive smart antenna processing device of the preferred embodiment. In each port, the initial weight vector 631-z, z=L·...N corresponding to the first port and the second port is used initially._sThe oversampled output 605 from the receiver 122 of the antenna element 103 is combined in the signal copy operation 607, and these initial weights are provided by the weight initialization operator 621. The resulting reproduction signal is corrected for time offset by a time offset corrector unit 609, which also samples/interpolates to generate a set of approximate baud-aligned samples (for iterations of the CM method) or basic baud-aligned The sample value (for the iteration of the decision-guided method) the sample value of the Bode alignment. The baud-aligned samples are fed to the SINR estimator 613, and its output is fed to the weight calculator and demodulator 615. The latter uses the wave-characteristic-aligned samples and/or the antenna signal 605 according to the method described in this article. The described method of the invention determines the reference signal and a set of weights. Because at least one iteration of the decision guidance method is used in the weight calculator and demodulator 615, the demodulated signal 617 is output. In this way, two demodulated signals of one subscriber unit are determined. With multiple ports, it can track the signal sent by any desired subscriber unit and any co-channel interferer at the same time. The adaptive method described further below can switch between any desired user and the interfering party in a fading environment. So using one port, N&amp; users are tracked at the same time, if any users

00819675.3 The first signal jumps from one port to another. This is possible in a fading environment. The output of the port is classified in the user classifier 623 to separate the desired user from any interfering parties and output correctly Ns demodulated signals 625.

There are other multi-port architectures that are well known, but they are not used in conjunction with the adaptive methods described here. See, for example, BG Agee's "Blind Separation and Acquisition of Communication Signals Using Multi-Target Constant Modulus Beamforming Device" [1989 IEEE Military Communications Conference (MILCOM 89) Vol. 2, No. 340-346, New York: IEEE, 1989] , Multi-port architecture for constant modulus method. Agee's method differs from the method described here in many ways, including, for example, 1) in the weight initialization method; 2) in what calculations are performed in each port. Agee's method commonly orthogonalizes the ownership vector in each iteration of the stage, and its computational cost is very high; and the best embodiment of the present invention allows each port to be independently adapted after the ports are initialized together; 3) The method of determining the weight vector is different. It should be noted that, because it is expected that computing power will be more and more easily available in the future, in an alternative embodiment, the weight vectors of each port are orthogonalized.

Best method for determining weight: multiple users FIG. 7 illustrates the best method for determining weight and output signal 625 through a flowchart. Initially, the eigenvector of the &amp;ζ matrix is used to perform the initial copy. The port marked "#1" is initialized with the eigenvector 631-1 corresponding to the largest singular value in step 703, the second port is initialized with the next eigenvector 631-2,..., and the first eigenvector is used 631 heart to initialize the second port. Ensure that these eigenvectors are linearly independent, and are generally the preferred values to start the calculation. Alternatively, any substantially independent initial weight vector can be used. For example, in an alternative embodiment, port #1 is initialized with the vector [10 0...0J, and port #2 is initialized with the vector [οιοο.,.ο]', and so on. In each port, after initialization, the method is executed in each port in this way for the single-user situation of the flowchart in FIG. 3. That is, step 305 is the copy operation performed first with the initial value. The resulting signal is subjected to time offset correction (including sampling/interpolation, if oversampling at the beginning) in step 307 to generate basic baud aligned samples, which are fed to the signal quality estimator, which In step 309, the SINR based on wave characteristics is estimated. In step 311, it is judged that the right

00819675.3 The first answer is whether to choose the CM method or the decision-making guidance method. If the SINR is lower than the predefined SINR threshold, then in step 313, the optimization based on the partial characteristic restoration method (preferably the CM method) is performed, and the method returns to step 305 to start the next step with the recently determined weight vector for this port Iteration. If the SINR is higher than the threshold, frequency offset correction is performed in step 315, and a single decision-guided adaptive iteration is performed in step 317. It should be pointed out that if the time offset correction in step 307 is only approximate, a more accurate correction will be required for the decision guidance method, and those skilled in the art should be aware of this modification. In the preferred embodiment, only one decision-guided iteration is performed. Or, perform more than one decision-guided iteration. When the weight is determined, because in the preferred embodiment, only a subset of samples in each burst is used, the weight vector finally determined for each port is used for the copy operation and demodulation steps performed on the entire burst. . The copy operation in the preferred embodiment includes the determination and correction of time and frequency offset and demodulation, and the demodulation preferably adopts the architecture shown in FIG. 4. The result corresponding to each port is the demodulated signal 617.

It should be pointed out that a common feature of the present invention in single-user and multi-user situations is that the right obtained by using the current burst data is used to determine the current burst signal. When the user unit moves around and is in other fading and low SINR environments, using the weight vector obtained from the previous burst may not get good results.

The final step is to classify these outputs to determine whether any output port has become locked by the interfering party. The PHS burst (for example, traffic channel burst) includes the fields of the payload, the unique word (UW) known to all user units, and the cyclic redundancy check (CRC) field for error detection. Other protocols include some different fields that can be used to determine whether a specific message comes from a specific subscriber unit, or whether it is for a specific base station. In the preferred embodiment, in order to determine the interference lock, the setting is to distinguish the required subscriber units that transmit waveforms that are valid for the system from the interfering subscriber units that also transmit waveforms that are valid for the system. That is, there are ways to define "valid" subscriber unit waveforms, for example, the waveform has a certain required data and modulation format, and is scrambled with a specific key for the subscriber unit. Similarly, the interference unit will include a certain way to define its own effectiveness, for example, the waveform has a certain required data and modulation format, and is different from the above-mentioned user unit.

00819675.3 scrambled with the specific key. In the best PHS implementation, one method for detecting interference lock includes monitoring the unique word (UW) and CRC at the same time. In particular, the data bits in each burst are scrambled with a bit pattern generated using the lower 9 bits of the cell station identification code (CSID). The 9-bit word used to encrypt the burst payload and its associated CRC is called the scrambling key. When designing a communication system, such as a cellular system, it is recommended to ensure that adjacent communication stations (base stations) each have different scrambling keys. It should be pointed out that in the PHS specification, a base station or communication station is called a cell station.

It is known that the basic interference lock detection method of the preferred embodiment includes the following steps: For a specific port, that is, a specific subscriber unit

1. Demodulate the received signal with the received right determined by the subscriber unit, and use the CSID-based key of the subscriber unit to descramble the burst payload;

2. Compare the received CRC with the CRC calculated according to the bit sequence after demodulation and descrambling of the burst payload;

3. Determine whether there is an obvious difference between the two, that is, indicating a transmission error or a wrong key, but UW does not display an error, and if the condition is met, the counter is triggered. In the case of including the right "tracking", if the condition is not met, it is assumed that the communication is not locked by interference, and the right (or the user unit space signature) used by the user unit is received as the "good" value of the user unit and saved (" Tracking"); and

4. If a certain number of consecutive bursts meet the conditions set forth in step 3, then according to the counter determined, it is judged that the port is confirmed to be interference locked.

In this PHS specification, when the co-channel users are all different spatial channels in the same regular channel, using the unique word and CRC to judge the interference lock will not work, because the CSID is the same for all subscriber units of the same regular channel of. For the case where the co-channel users are spatial channels of the same regular channel, the judgment of interference lock can be achieved by maintaining the spatial signature history of these co-channel users.

Once the output is classified, the result is a set of output signals from each port.

Apparatus FIG. 9 shows a block diagram of an apparatus implementing an aspect of the present invention. Used for receiving special

00819675.3 The device for determining the weight vector of the specific signal transmitted by the predetermined user unit includes: an initialization device 902, which is used to initialize with a first initial vector value; and a first iterative device 905, which is used to modify the weight vector in an iterative manner according to the first iterative method The first iterative method minimizes the first cost function. It is an iterative weight determination method with good convergence characteristics, preferably the constant modulus method implemented as described above. The device also includes a second iterative device 907, It is used to modify the weight vector in an iterative manner according to the second adaptive method that minimizes the second cost function. It is a fast-convergent iterative weight determination method, preferably the above-mentioned decision guidance method. The initialization device 902, the first iterative device 905, and the second iterative device 907 are under the control of the control device 911, and the control device 911 is programmed to activate the first iterative device 905 from the first initial vector value provided by the initialization device 902 Until the switching criterion is met, the final weight vector after the last iteration of the first adaptive method is the second vector value; and starting from the second vector value, the second iterative device 907 is activated to determine the weight vector 909 . The weight vector 909 is used by the spatial processor and demodulator 915 to generate a demodulated signal when instructed by the controller 911, which uses the signal received on the antenna array 103 by the receiver 122. Various iterative methods include determining the replicated signal. The device preferably includes an SINR estimator 913 to use the weight vector determined by the first iterative device 905 to estimate the replicated SINR in the replicated signal of the first iterative device. The switching criterion is preferably that the SINR estimated value exceeds the SINR threshold.

The right determination device preferably includes at least one digital signal processor (DSP) device in the base station, and each component 902, 905, 907, 909, 911, 913, and 915 is preferably implemented in the form of one or more DSP programs. Those skilled in the art should understand that, without departing from the spirit and scope of the present invention, skilled practitioners can make many modifications to the above methods and devices. For example, a communication station that implements this method can use one of a variety of protocols. In addition, multiple architectures of these stations are also feasible. Many further changes are feasible. The true spirit and scope of the present invention should be limited only by what is stated in the claims.

00819675.3