CN1234926A - Signal processor employing spectral property of signals - Google Patents

Abstract

A method and apparatus for separating and removing distortion from interfering co-channel signals and suppressing adjacent-channel interfering signals of the Gaussian Minimum-Shift Keyed (GMSK) or other MSK type with filtering structures that exploit the cyclostationarity of the received GMSK or other MSK signals in order to accommodate a greater number (or the same number, but with greater quality) of transmitted signals received by one or more antennas than can be accommodated by existing filters. The parameters in these filtering structures are adapted by either of two adaptation apparatus that exploit both the known training sequence that is transmitted in most wireless communications systems, and the constant modulus property exhibited by each of the transmitted GMSK or other MSK signals.

Description

Utilize the signal handling equipment of spectral property of signals

The present invention relates generally to utilize GMSK (Guassian Minimum Shift Keying) (GMSK) or other minimum shift keying (MSK) signal to reduce interference in the wireless communication system, and relate more specifically to a kind of method and apparatus, being used for (ⅰ) is suppressed at that receive at the communication control processor place and occupies interference signal identical with one or more desired signals or nearby frequency bands, (ⅱ) if desired, the a plurality of required signal of communication that the communication control processor place is received is separated, the distortion that reaches one or more required signals of communication that (ⅲ) will influence the reception of communication control processor place is proofreaied and correct, and for example propagates the distortion that channel caused by multipath radio frequency (RF).

Background of the present invention is divided into two big classes: (ⅰ) about being used to finish the work of the filter structure that disturbs inhibition, Signal Separation and distortion correction, and (ⅱ) work of relevant adaptive algorithm, this algorithm adjust in these filter structures coefficient with allow they move, can take with the peculiar signal environment of personal wireless communications system in move.Yet this work is different with the present invention basically.

A. filter structure

Except that single-sensor FRESH filter and joint maximum likelihood sequential estimator, all other be used for same frequency band and sequential receive at interval a plurality of GMSK or msk signal system the work on hand of applicable interference inhibition and Signal Separation all require to use a plurality of transducers (antenna).Another requirement of this work is that the spatial character of any two signals of receiving of a plurality of transducers must be sufficiently different; So require to be not being met (or satisfying limpingly), then filter structure can't suppress to disturb and separate desired signal (or accomplish this point with poor quality).This work comprises that all belong to the filter structure of linear time invariant (LTI) space time filter class; This type of filter is by following true decision: promptly the amplitude of their the available a plurality of signal of sensor of output is demarcated, phase shift and time shift form and ask mould with coming mathematics.Obviously, this type of filter in fact characterizes the filter structure of developing before all that is used to handle the problem that solved by the present invention (and be similar to its problem).

As for single-sensor FRESH filter, do not have one (ⅰ) to discuss in the work on hand or describe the FRESH filtering application in GMSK or other msk signal class, (ⅱ) discussion or description are to the application of GSM or dect standard, (ⅲ) frequency displacement is used in suggestion, this frequency displacement equal to add deduct half of signal bit rate, or equal add deduct half of bit rate of twice carrier frequency, (ⅳ) show one with have same carrier and bit rate and do not comprise a plurality of GMSK signals (for having the complex packet winding thread that zero carrier is offset) in non-conjugated frequency displacement path relevant 3 the tunnel through the LCL-FRESH structure, (ⅴ) show and 2 relevant paths of adjacent channel GMSK interference, 2 input LCL-FRESH structures, (ⅵ) show the 3 paths 3 input LCL-FRESH structures relevant with adjacent channel GMSK interference, (ⅶ) show a multisensor 3 path LCL-FRESH structures of each same form of 2 paths of each the transducer output comprise its all 3 path LCL-FRESH filter output variables summations or 3 path LCL-FRESH structures, (ⅷ) description is used for adaptive any filter to separate the blind dark method of any part of GMSK or msk signal, (ⅸ) describe the auxiliary norm of the training that is used for adaptive any class signal and figure method (CMA) method, (ⅹ) separability of using any FRESH filtering method to show any GMSK or msk signal, (ⅹ ⅰ) describes and uses FRESH filtering to provide the theory of the GMSK of signal separability to compose redundant character, or (ⅹ ⅱ) shows the redundant character of spectrum of the measurement of GMSK signal.In addition, single-sensor FRESH filter has a very big restriction that can be surmounted by the present invention: different with the present invention, how it does not hold a plurality of transducers yet, thereby yet just can not suppress and/or separate the GMSK or the msk signal of and 2 adjacent channels co-channel more than 2.

Though the joint maximum likelihood sequential estimator of operating with single-sensor (JMLSE) has the ability inhibition and/or separates more than 2 GMSK or msk signal in theory, when number of signals surpassed 2, the quality of institute's separation signal in fact necessarily reduced.In addition, the computational complexity of JMLSE is very big, therefore need not expensive hardware, and it is unpractical not using a large amount of electric power and will being implemented in real time.Surpass 1 and the number of signals that needs jointly to separate when increasing when the number of sensors held increases, these shortcomings become even more serious (rising with index percent).

Filter structure of the present invention does not possess the serious enforcement shortcoming of JMLSE; They advantageously hold single or multiple transducers; Even in the time can't and/or separating one or more pairs of signal with the inhibition of LTI space time filter, they also can suppress and/or separate these signals.More particularly, filter structure providing capability of the present invention be used to suppress interference signal and separate desired signal and their distortion of correction, and any existing filter structure can't provide these abilities.

B. adaptation method

Among the present invention and be used to disturb other device of inhibition, Signal Separation and distortion correction and the used filter structure of equipment by selective filter parameter or coefficient value and Be Controlled.In some embodiment of these filter structures, these coefficients can comprise be used for being the gain and the phase place of unlike signal weighting before summation.In common wireless communication systems, the value of these coefficients must appropriate change, because of the space-time characterisation of the signal received changes in time.Also promptly, the coefficient value of these filter structures must be selected adaptively.

Except that the present invention, have two class methods to be used for selecting adaptively these coefficient values: (ⅰ) conventional method, it depends on the knowledge of contained training sequence in the desired signal uniquely, and this knowledge is generally used for selecting according to the minimum variance criterion initial setting of coefficient value; Reach (ⅱ) blind dark method, it does not use contained training sequence, opposite one or more characteristics of but utilizing interference or desired signal.

Use the conventional method of known training signal to say it is well-known, and in the open source information of relevant Adaptive signal processing and adaptive filtering, description is arranged for the people who is familiar with signal processing technology.Yet conventional method is limited in their application, because of the coefficient value quantity that they can be successfully adaptive must be less than the quantity of known sample value in (just must be significantly smaller than for having obtained performance) contained training sequence.This known number of sam-pies by the wireless communication system defined (for example, for all signals among the GSM that transmits the sound traffic, in discrete time was for example implemented, this quantity equaled 26 and multiply by used every effective independent sample number (being generally 2) in the digitlization of received signal continuous time).About problem involved in the present invention, wherein filter structure may have a large amount of adjustable integral coefficients potentially, and the known number of sam-pies in the contained training sequence may can't provide the initial setting of usefulness inadequately.Initialized in this way filter structure can't be provided at enough high-quality of desired signal useful in the communication system and estimate.

On the contrary, blind dark method can use the received signal under the constant approx space-time characterisation whole records (be the incidence angle of the multi-spread path of many signals and carrier wave, the relative delay, with relative phase and transmit the position mutually).This has overcome the restriction of conventional method on filter coefficient value that can be adaptive.Yet different blind dark methods have other critical defect, thereby can't successfully move in many instances, and these shortcomings comprise that the present invention points out.Especially, norm is figured these class methods of method (CMA) and method and is required than the also much longer data record that can accomplish so that converge to the high-quality solution.In addition, whether they can converge to solution does not guarantee, this causes producing between interference signal (should give inhibitions) and desired signal (answer in the self-interference separation and be separated from each other) obscures.Another kind of blind dark method depends on the estimation of incident direction to received signal; Use these to estimate in a question blank (by array bunch, it is the nominal data table of multisensor receiver), to find out the coefficient of LTI space time filter then.Yet when there being a large amount of multipaths, during distinctive destructive factor, the use result of these so-called direction method for searching is very poor in the interested wireless communication system of the present invention for these three kinds in unknown number of signals and unknown interference and noise territory.In addition, the direction search method is not provided as adaptive filter structure of the present invention and the parameter estimation type of needs.

Adaptation method of the present invention with novel manner use simultaneously the norm of contained training sequence and desired signal count characteristic the two, and by doing the bad aspect that overcomes conventional method and blind dark method like this.Also promptly, adaptation method of the present invention is selected coefficient value in the filter structure, all values for example of the present invention, thereby provide existing adaptation method the ability that can't provide.

The method and apparatus of distortion is separated and corrected to the desired signal that the invention provides a kind of interference that is used for suppressing bad and the GMSK (Guassian Minimum Shift Keying) (GMSK) that will occur at for example wireless communication system or other minimum shift keying (MSK) type, or it is according to digital european cordless telecommunications (DECT) standard or according to global system for mobile communications (GSM) standard operation, consequently improves the quality and the quantity of the service that wireless communication system provides of using GMSK or other msk signal.

Wireless Telecom Equipment manufacturer and the radio communication service supplier and/or wish who wish to improve the users' that offer their systems communication quality increases the number of users (for example increasing income) of their systems in commercial operational system people will appreciate that, but not uniquely, application of the present invention.

When distance is reused by for example reducing frequency in the base station, spreading range, may be in mobile radio station reusable frequency and give adjacent channel with the user more frequently, thereby when supporting the more users of fixed Frequency Assignment, quantity of service is improved.When for example the request of calling out to Consultation Center's (being called the base station from now on) proposition startup as the user less is rejected (promptly reducing blocking probability), when mobile subscriber's the calling of just carrying out is not non-normal termination (promptly reducing to call out the probability that rolls off the production line), and the voice message intelligibility (is that calling both sides is heard less noise better in calling out, disturb, improper noiseless etc.) time, service quality improves.

The present invention relies on the method and apparatus of finishing following function to solve these general considerationss: (1) suppresses communication control processor interference signal that receive and that occupy identical with one or more desired signals (or adjacent) frequency band, (2) (if necessary,) a plurality of required signal of communication that communication control processor is received separated, and the distortion (for example propagating the distortion that channel causes by multipath radio frequency (RF)) of one or more required signals of communication of receiving of (3) correct influences communication control processor.

Therefore, an object of the present invention is to provide an array of one or more antennas, the output variable of these antenna is by conventional RF front-end processing, suppress to disturb by one then, separate one or more desired signals and the genuine filter structure of correction husband and handle to obtain the high-quality estimation of desired signal.

Another object of the present invention provides a class filter structure, and they can suppress interference signal, separates one or more desired signals and proofreaies and correct distortion in these desired signals.

Another purpose of the present invention provides a class filter structure, they can by conventional RF front end with the output variable down-conversion of any amount of antenna after to the operation of these output variables, comprising the special circumstances of having only an antenna.

A further object of the invention provides a linearity-conjugation-linearity (LCL) frequency displacement (FRESH) filter, its uses the AD HOC of frequency displacement and conjugation so that utilize cyclostationarity as GMSK and other msk signal characteristic, thereby provide required interference to suppress Signal Separation and distortion correction capabilities.

Another aspect of the present invention provides linearity-conjugation-linearity constant during the multicycle (LCL-PTV) filter, its use conjugation and the time constant filter coefficient AD HOC so that utilize cyclostationarity as GMSK and other msk signal characteristic, provide required interference to suppress Signal Separation and distortion correction capabilities from face.

A further object of the invention provides a LCL-FRESH mark spaced equalizers (LCL-FRESH-FSE), it to the data manipulation that is higher than bit rate sampling with bit rate sampling output variable is provided and use frequency displacement and the AD HOC of conjugation to use cyclostationarity as GMSK and other msk signal characteristic, thereby provide required interference to suppress Signal Separation and distortion correction capabilities.

Another object of the present invention provides a LCL-FSE filter, it to be higher than bit rate sampling and (by 1/4th bit rates) data manipulation frequency displacement and use data that FSE handles and conjugation data sum to use cyclostationarity as GMSK and other msk signal characteristic, thereby provide required interference to suppress Signal Separation and distortion correction capabilities.

Another purpose of the present invention provides a reality-void (RI-FSE) filter, and it provides the iteration embodiment of effective LCL-FSE filter in the calculating.

Another purpose of the present invention provides a LCL filter, its bit rate sampling and (by 1/4th bit rates) data manipulation frequency displacement and use the linear filtering data and conjugation data sum so that use cyclostationarity as GMSK and other msk signal characteristic, thereby provide required interference to suppress Signal Separation and distortion correction capabilities.

Another purpose of the present invention provides a reality-void (RI) filter, and it provides the iteration embodiment of effective LCL filter in the calculating.

Another purpose of the present invention provides class methods that are used in the adaptive adjustable integral coefficient of any filter structure of the present invention and suppresses interference, the function of separating one or more desired signals and correcting distortion so that this filter can suitably be finished.

Another aspect of the present invention provides a kind of device, and its uses the known training sequence be contained in the desired signal and desired signal to have norm and counts this knowledge and come adaptive adjustable integral coefficient, thereby provides the norm number (TACM) of training augmentation adaptive.

A further object of the invention provides a device, its uses the known training sequence be contained in the desired signal with adaptive adjustable integral coefficient, thereby that the adaptive norm number (TCCM) of training constraint that provides of this knowledge is provided is adaptive so that constraint uses desired signal to have norm.

These general aspects and their details make the present invention have very big ability that extensive use is provided in wireless communication system.Ability provided by the present invention can be used for improving the quality and the quantity of the communication service that type systematic thus provides.Use the present invention suppress in special frequency band and the timeslice (jointly being called channel) with space cell in cause by other users in identical or the adjacent channel with identical or adjacent space unit in other user interference and the distortion correcting signal that cause, can improve service quality, and channel redistributed strengthen with the interference signal of having a mind to make each channel and unit but can separate by the present invention, thereby can hold more users, can increase quantity of service.

Because the present invention only uses a transducer, energy separates a plurality of spectrum overlap signals of certain type and can proofread and correct even the decline of serious frequency selectivity and other distortion, and, just may realize improvement of the present invention because available a plurality of transducers increase the quantity and the number of sensors of this class signal that can be separated by the present invention pro rata.In the beacon nest communication system of using the base station, so channel distribution (but be not necessary) so that the interference of introducing only in upstream channel (from mobile device to the base station), occur, thereby only need be in base station receiver and needn't in mobile receiver, implement the present invention.The base station that the present invention also can be used for only using single receive antenna or use a plurality of antennas (or be used for mobile device, or be used for both).

With reference to the object lesson of GMSK and other msk signal, the GMSK between 2 and 4 in the single channel and other msk signal can hold with an antenna, depend on that how many signals the user sends and send how many signals in adjacent channel in the same channel.More generally say, can hold signal between 2M and the 4M with the M number of antennas.

The lower part of specification will be introduced other purpose of the present invention and advantage, and wherein the purpose of Xiang Ximiaoshuing is whole the preferred embodiments of the present invention and it is not applied restriction openly.

With reference to following accompanying drawing that is used to explain purpose, can more fully understand the present invention, in the accompanying drawing:

Fig. 1 is the figure that shows the cross-coherence function of a sample value GMSK signal and its conjugation and frequency displacement form.

Fig. 2 A is the figure that shows the spectrum content of GMSK (0.3) signal x (t).

Fig. 2 B is the conjugation and the frequency displacement form of signal shown in the displayed map 2A $y_1\left(t\right)=x^{*}\left(t\right)e^{-\mathrm{\π}{f}_{d}t}$ The figure of spectrum content.

Fig. 2 C is y shown in x shown in the displayed map 2A and Fig. 2 B ₁Ask the figure of the spectrum density of correlation.

Fig. 3 A is the figure that shows the spectrum content of GMSK (0.3) signal x (t).

Fig. 3 B is the conjugation and the frequency displacement form of signal shown in the displayed map 3A $y_2\left(t\right)=x^{*}\left(t\right)e^{j{\mathrm{\πf}}_{b}t}$ The figure of spectrum content.

Fig. 3 C is y shown in x shown in the displayed map 3A and Fig. 3 B ₂Ask the figure of the spectrum density of correlation.

Fig. 4 is one and is used to finish the interference inhibition of GMSK or other msk signal and/or the functional block diagram according to 3 path LCL-FRESH filters of the present invention of co-channel separation and/or distortion correction.

Fig. 5 is one and is used to finish the interference inhibition of GMSK or other msk signal and/or the functional block diagram according to LCL-PTV filter of the present invention of co-channel separation and/or distortion correction.

Fig. 6 is one and is used to finish the interference inhibition of GMSK or other msk signal and/or the functional block diagram according to " Class1 " of the present invention LCL-FRESH-FSE filter of co-channel separation and/or distortion correction.

Fig. 7 is one and is used to finish the interference inhibition of GMSK or other msk signal and/or the functional block diagram according to " type 2 " of the present invention LCL-FRESH-FSE filter of co-channel separation and/or distortion correction.

Fig. 8 is one and is used to finish the interference inhibition of GMSK or other msk signal and/or the functional block diagram according to " Class1 " of the present invention LCL-FSE filter of co-channel separation and/or distortion correction.

Fig. 9 is one and is used to finish the interference inhibition of GMSK or other msk signal and/or the functional block diagram according to " type 2 " of the present invention LCL-FSE filter of co-channel separation and/or distortion correction.

Figure 10 is one and is used to finish the interference inhibition of GMSK or other msk signal and/or the functional block diagram according to " Class1 " of the present invention RI-FSE filter of co-channel separation and/or distortion correction.

Figure 11 is one and is used to finish the interference inhibition of GMSK or other msk signal and/or the functional block diagram according to " type 2 " of the present invention RI-FSE filter of co-channel separation and/or distortion correction.

Figure 12 is one and is used to finish the interference inhibition of GMSK or other msk signal and/or the functional block diagram according to LCL filter of the present invention of co-channel separation and/or distortion correction.

Figure 13 is that an interference that is used to finish GMSK or other msk signal suppresses and/or co-channel separation and or the functional block diagram according to RI filter of the present invention of distortion correction.

Figure 14 is the interference inhibition and/or the functional block diagram according to 5 path LCL-FRESH filters of the present invention co-channel separation and/or distortion correction and that improve the counteracting that adjacent channel is disturbed that are used to finish GMSK or other msk signal.

Figure 15 is one and is used to offset the adjacent functional block diagram according to 2 inputs, 2 path LCL-FRESH filters of the present invention that channel disturbs of going up.

Figure 16 is one and is used to offset the functional block diagram according to 2 inputs, 2 path LCL-FRESH filters of the present invention that adjacent channel down disturbs.

Figure 17 is one and is used for offsetting the functional block diagram according to 3 inputs, 3 path LCL-FRESH filters of the present invention that adjacent upper and lower two channels disturb.

Figure 18 is a functional block diagram according to joint spatial-temporal multisensor FRESH filter of the present invention.

Figure 19 is a functional block diagram according to space-time resolution process device of the present invention.

Figure 20 is the data structure sequential chart of timeslice in the gsm system, shows the quantity and the position of each class position.

Figure 21 is the period of time T that shows timeslice shown in Figure 20 ₀And T ₁The data structure sequential chart.

Figure 22 is the functional block diagram according to receiver device of the present invention.

Figure 23 is the functional block diagram of relevant M antenna collecting device piece embodiment shown in Figure 22.

Figure 24 is the functional block diagram of the piece of filter apparatus shown in Figure 22 embodiment.

Figure 25 is the frequency displacement of filter shown in Figure 24, the functional block diagram of time shift and polarization switch device block embodiment, and wherein filter configuration is 3 path LCL-FRESH filters.

Figure 26 is the frequency displacement of filter shown in Figure 24, the functional block diagram of time shift and polarization switch device block embodiment, and wherein filter configuration is a Class1 LCL-FRESH-FSE filter.

Figure 27 is frequency displacement shown in Figure 26, the iteration embodiment of time shift and polarization switch equipment, and wherein filter configuration is a type 2LCL-FRESH-FSE filter.

Figure 28 is the frequency displacement of filter shown in Figure 24, the functional block diagram of time shift and polarization switch device block embodiment, and wherein filter configuration is a Class1 LCL-FSE filter.

Figure 29 is frequency displacement shown in Figure 28, the iteration embodiment of time shift and polarization switch equipment, and wherein filter configuration is a type 2LCL-FSE filter.

Figure 30 is the frequency displacement of filter shown in Figure 24, the functional block diagram of time shift and polarization switch device block embodiment, and wherein filter configuration is a Class1 RI-FSE filter.

Figure 31 is the frequency displacement of filter shown in Figure 24, the functional block diagram of time shift and polarization switch device block embodiment, and wherein filter configuration is the LCL filter.

Figure 32 is the frequency displacement of filter shown in Figure 24, the functional block diagram of time shift and polarization switch device block embodiment, and wherein filter configuration is the RI filter.

Figure 33 is frequency displacement shown in explaination Figure 25 to Figure 32, the functional block diagram of the polarization switch part of time shift and polarization switch equipment.

Figure 34 is frequency displacement shown in Figure 25 to Figure 32, the functional block diagram of tap time delay (TDL) the device block embodiment of time shift and polarization switch equipment, wherein the input and output vector can be real number or the plural number.

Figure 35 is the functional block diagram according to TACM adaptability equipment of the present invention.

Figure 36 is the functional block diagram according to TCCM adaptability equipment of the present invention.

More specifically with reference to accompanying drawing, the present invention obtains explaination, wherein identical reference numerals same parts generally referring to figs. 1 through the demonstration of Figure 36.Should be understood that under the situation that does not deviate from basic conception disclosed herein this equipment can be different on step and their order in different and this method on configuration and the component detail.

A. the theoretical foundation of filter structure and they

Filter structure of the present invention is used GMSK and 100% shown spectrum redundancy of other msk signal.Also promptly, these filter structures are based on the following fact: the information that transmits in one frequency band of GMSK or other msk signal also transmits in another time frequency band.In fact, for any GMSK or other msk signal, the waveform in any two so relevant inferior frequency bands almost is relevant fully.Therefore, this has composed redundant augmentation more extensive spatial redundancy of understanding and using, wherein given source is caused and fully is correlated with by the waveform that two different antennae in the array are received.

Do not have a plurality of antennas as one be not used in the receiver that suitably makes up the required space filtering structure of their output variables and can not get owing to the redundant well-known performance that obtains in application space is the same, one does not have the receiver of filter structure of the present invention to can not get because the outstanding properties that the spectrum of use redundancy obtains.

By the application space redundant and spectral redundancy (also claiming to compose relevant) both, filter structure of the present invention provides the processing gain level of twice LTI spatio-temporal filtering structure at least.For example, the M aerial array can use these filter structures to isolate as many as 2M to 4M GMSK or other msk signal, perhaps can isolate M GMSK or other msk signal that as many as can obtain more traditionally, but can accomplish this point with reliability and the quality more much higher than conventional spatial filter.Relevant and with reference to the spectrum of GMSK or other msk signal with reference to the relation between correlation (no matter its origin or type how) and the linear filtering, soluble these abilities.

B. the relation between correlation and the linear filtering

Though the relation between correlation and the linear filtering is basic, also not necessarily understand it even be familiar with the people of signal processing technology.Therefore, in order to describe purpose of the present invention, we use a simple but abstract class recently to explain this relation: as x (t)=a (t)+b (t), and y (t)=c (t)+d (t), wherein a (t) and c (t) are height correlations, and wherein b (t) and d (t) are height correlations, then x (t) can provide one or more signal a (t) with the appropriate linear combination (promptly jointly using the output variable of the linear filtering structure of selecting rightly of x (t) and y (t)) of y (t), b (t), the high-quality of c (t) and d (t) is estimated.

The spectrum correlation of C.GMSK and other msk signal

How to make this performance become possibility for understanding, and which kind of filter structure of understanding can be applicable to GMSK and other msk signal, the necessary spectrum correlation properties of understanding GMSK and other msk signal.For this purpose, these characteristics can explain with the cross-coherence function best that it is a frequency domain crosscorrelation coefficient.Specifically, make x (t) that expression has zero carrier shift and bit rate f _bThe GMSK signal the complex-envelope line and make y (t) be conjugation and the frequency displacement form of x (t), wherein frequency displacement is f _b/ 2 or-f _b/ 2. $y_1\left(t\right)=x^{*}\left(t\right)e^{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="A9618046500372.png,A9618046500381.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;}(f_b/2)t}----\left(1\right)$ $y_2\left(t\right)=x^{*}\left(t\right)e^{+\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="A9618046500372.png,A9618046500381.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;}(f_b/2)t}-----\left(2\right)$ Y (t)=y wherein ₁(t) or y (t)=y ₂(t).We consider the cross-coherence function C _Xy(f) define by following formula $C_{\mathrm{xy}}\left(f\right)=\frac{S_{\mathrm{xy}}\left(f\right)}{\sqrt{S_{\mathrm{xx}}\left(f\right)S_{\mathrm{yy}}\left(f\right)}}----\left(3\right)$ S wherein _Xy, (f) cross-spectral density between expression x (t) and the y (t), and S _Xx(f) and S _Yy(f) represent the power spectral density of x (t) and y (t) respectively.Fig. 1 shows the amplitude of GMSK signal | C _Xy(f) | with the relation of frequency f, wherein solid line representative | C _Xy1(f) | and the dotted line representative | C _Xy2(f) |, and distribute to the whole channel of signal (100kHz to+100kHz) on clearly very near 1.

For | C _Xy(f) | any f of 1 no better than, the arrowband component of x at frequency f place (t) and y (t) is almost relevant fully.For our definition of x (t) and y (t), this is equivalent to below the argumentation: the arrowband component of the x at f place (t) with or f+f _b/ 2 places or f-f _bThe arrowband component of the x* at/2 places (t) is almost completely relevant.With reference to Fig. 2 A to Fig. 2 c, sequence number is-6 ,-5 among Fig. 2 c ... ,-1,0 arrowband component | C _Xy1(f) | 1 this fact means among arrowband component that every pair of sequence number obtaining in x (t) among Fig. 2 A is identical and Fig. 2 B no better than $y_1\left(t\right)=x^{*}\left(t\right)e^{-j{\mathrm{\&pi;f}}_{b}t}$ Almost be relevant fully.X (t) and y for identical f value place ₁(t) component right continuum in arrowband in, this is right.Therefore, and y in these components among the x (t) ₁(t) combination of the proper linearity of corresponding one-component can provide that the proper linearity combination of the signal that received of the many antennas of twice is identical at least with respect to the used antenna number when receiving x (t) of the processing gain level of other signal, interference and noise in.Analogue be applied to x (t) with $y_2\left(t\right)=x^{*}\left(t\right)e^{j{\mathrm{\&pi;f}}_{b}t}$ Linear combination.Therefore, one is used x (t), y ₁(t) and y ₂It is identical with respect to the proper linearity combination of the signal that the many antennas of used antenna number twice are received when receiving x (t) of the processing gain level of other signal, interference and noise that the filter structure of linear combination (t) can provide.

According to these facts of the spectrum correlation properties of relevant GMSK and other msk signal, quote two class certain interference schemes herein as an example, be used for explaining but do not limit the applicability of the practical problem filter structure of the present invention of wireless communication system.For simplicity, in two class schemes, all only use individual antenna; When used antenna amount increased, the ability of the filter structure of advising increased pro rata.

Co-channel interference suppresses

Individual antenna receives a required GMSK signal and one and has same carrier and the equal co-channel GMSK signal of signal power.Constant filter structure recovers during neither one signal useable linear.Yet, with reference to Fig. 2 A to Fig. 2 c, among the x (t) each sequence number be-6 to 0 component (and the component in the continuum of-150kHz to 0kHz) can with its at y ₁(t) respective component in makes up with the clean estimation that produces a desired signal linearly and suppresses co-channel interference.Similarly, with reference to Fig. 3 A to Fig. 3 C, each sequence number is that 0 to 6 component (and the component in 0 to 150kHz continuum) can make up with the clean estimation that produces a desired signal linearly with its respective component in y2 (t) and suppresses co-channel interference among the x (t).Therefore, use y ₁(t) and y ₂(t) both, can be on the whole frequency band to+150kHz from-150kHz, also can be as required only on the GSM channel of the distribution to+100kHz, recover desired signal from-100kHz.In this way, the filter structure among the present invention even only with an antenna time, also can suppress the co-channel interference signal of GMSK.

Adjacent channel disturbs and suppresses

Individual antenna receives a required GMSK signal and the adjacent channel GMSK signal that may have much bigger signal power.Specifically, establishing the carrier frequency less than the desired signal carrier frequency that adjacent channel signals has is 200kHz (by GSM allocation of carriers agreement).Because spectrum is leaked, the adjacent channel interference source degenerates desired signal, and it can't be recovered by the linear time invariant filter structure.Yet, with reference to Fig. 3 A to Fig. 3 C, adjacent channel disturb spectrum in (now being converted to the zero carrier skew) leak each sequence number that enters in the desired signal be 4 to 8 component (and the component in the continuum of 100kHz to 200kHz) can with its at y ₁(t) respective component makes up linearly to offset the influence that spectrum is leaked.Similarly, its carrier frequency can be offset for the adjacent channel interference source of the 200kHz on the desired signal carrier frequency.In this way, even only use an antenna, the filter structure among the present invention also can suppress two GMSK adjacent channel interference signals.

Distorted signals is proofreaied and correct

When the signal of emission by undesirable channel for example during the multipath transmisstion channel propagation, received signal may be the serious forms of distortion that is transmitted, and comprises for example frequency selective fading (promptly getting deeply stuck in the spectrum null value).Yet, with reference to Fig. 2 A to 2C, the sequence number that is seriously weakened (weaken when in so much consequently they being amplified and can greatly increase noise level) among the x (t) by frequency selective fading is-6,-5, the all available y of in-1,0 the arrowband component any one (and the component in the continuum of-150kHz to 0kHz) ₁(t) component that corresponding phase modulation rightly and amplitude are demarcated in substitutes, and composes zero point thereby remove.Similarly, with reference to Fig. 3 A to Fig. 3 C, the spectrum in the frequency band to+150kHz from 0kHz can be removed zero point.

D. specific filter structure

Describe according to different filter structures of the present invention referring now to Fig. 4 to Figure 17.Herein in Chan Shi the filter structure, constant filter when all inside have only the piece of a letter to represent normal linearity (ignoring the time variation that only causes) by adaptability, have a letter and back and represent linearly periodically time varing filter with those pieces of (t), and those pieces with a letter and FSE represent the mark spaced equalizers (filter) of linear time invariant, and all these are the known symbols that use in the signal processing technology.In addition, in have the piece of symbol () * to represent to ask conjugator, in have the piece of " Re " to represent real part selector (extractor), and in have the piece of " Im " to represent imaginary part selector (extractor), each is operated complex signal among them, and is also referred to as polarization switch at this.These elements are to be familiar with perhaps not so will describing herein known to the people of normal signal treatment technology.

The LCL-FRESH filter

The front is relevant | C _Xy(f) | observation be in 3 path time domain linear-conjugation-linearity (LCL) frequency shift filters (FRESH filter) 10 according to the present invention, to obtain considering promoting the relevant effect medium-altitude or almost completely of linear filtering structure, the structure that this filter 10 is used for GMSK and other msk signal obtains explanation with the block diagram form at Fig. 4.When the signal x (t) that is used for being received by the individual antenna place goes up, separable two the co-channel GMSK of two copies or other msk signal of this structure or other filter structure of describing below and/or proofread and correct their distortion.3 path LCL-FRESH filters are applicable to the data with the speed sampling that is higher than bit rate.Therefore, 3 path LCL-FRESH filter output variables must be with the bit rate sampling before processing, and for example the GMSK demodulator of being operated by the bit rate sampled data is sampled to recover bit stream.

The LCL-PTV filter

As the iterative scheme to 3 path LCL-FRESH filters, the frequency displacement of two complementations With

Filter g with associated _-And g ₊The also available single time varing filter g of effect (t) finish.Become (LCL-PTV) filter 12 when resulting structures is called linearity-conjugation-linear period herein, it is shown among Fig. 5 with the block diagram form.The time varying impulse response that filter g (t) becomes when having according to the possible cycle in the possible cyclo-stationary statistical property of received signal.In some applications, when existence was shone upon one to one between LCL-PTV filter and 3 path LCL-FRESH filters, making time varing filter g (t) coefficient was suitable along with frequency f b changes.In other is used, for example when signal the sheet border not well to punctual and the time leaf length when being the non-integer multiple of bit period (for example in GSM), sheet border when one or more received signals run into, filter g (t) coefficient just must change.For the data of sampling with the speed that is higher than bit rate, the LCL-PTV filter is suitable.Therefore, LCL-PTV filter output variable must be sampled with bit rate before by the GMSK demodulator processes that for example is operated the bit rate sampled data, so that recover bit stream.

The LCL-FSBSH-FSE filter

In 3 path LCL-FRESH filters and LCL-PTV filter, when the filter output variable is sampled with bit rate before being handled by demodulator (estimation of the bit stream that it transmits in providing for desired signal) usually, the sample rate of filter output variable surpasses bit rate, and (for the linear filter of handling GMSK, it is bit rate f usually _bInteger multiple, 2f for example _b).Therefore, the iterative scheme of 3 path LCL-FRESH filters be under the situation that does not change the filter ability as much as possible with this bit rate sampling operation after filtration wave structure move so that eliminate redundant operation backward.Explain the iteration embodiment of gained filter structure 14 (Class1) and 16 (type 2) among Fig. 6 and Fig. 7 with the block diagram form respectively, claim that herein they are LCL-FRESH mark spaced equalizers (LCL-FRESH-FSE).The remarkable difference of constant filter is specifically using of its conjugated signal path and to the special use of its frequency displacement when this filter and LCL, thereby suitably uses the cyclostationarity of GMSK or other msk signal.The bit rate sampled signal is estimated Must be by demodulator processes to recover bit stream.

LCL-FSE and RI-FSE filter

Can on the LCL-FRESH-FSE filter, realize the direct manipulation with obtain one wherein each bit rate sampler output variable by j ^-kThe filter structure of frequency displacement.This frequency shift signal is divided into two-way then: go up directly to be reached down by the LTI filter process and directly ask conjugate and processing by another LTI filter.Also promptly, follow each bit rate sampler following footpath afterwards and no longer comprise (1) ^kFrequency displacement.Yet output final stage adder must be by j ^kFrequency displacement.The iteration embodiment that shows gained filter structure 18 (Class1) and 20 (type 2) among Fig. 8 and Fig. 9 respectively.This filter structure is called LCL-FSE filter rather than LCL-FRESH-FSE herein, is different with the frequency displacement in any other path because there is not the frequency displacement of a paths.

Can under the situation that does not reduce performance, simplify the LCL-FSE filter: two paths and the j that notice each bit rate sampler output ^-kFrequency shifter forms a LCL filter, it can be realized equivalently by following measure, footpath in the change reaches footpath under the change so that in operation (rather than operating) on the imaginary part so that in operation (rather than operating) on the real part on the conjugate complex number signal on whole complex signal.Because this structure to replacing two LCL filters, so resulting structures is called the RI-FSE filter, and may be interpreted as the arithmetically equivalent iteration embodiment of LCL-FSE filter with two reality-void (RI) filter.The iteration embodiment 22 (Class1) and 24 (type 2) that shows this filter structure among Figure 10 and Figure 11 respectively.

LCL and RI filter

LCL is relevant with the RI-FSE filter with LCL-FSE with the RI filter, but the bit rate sampled data is operated.This is to GMSK and be possible to some other msk signal, because bit rate is greater than the shared RF bandwidth of signal, so this bit rate satisfies the Nyquist sampling criterion of this signal.

Figure 12 shows the LCL filter 26 as the simplification result of the LCL-FSE filter of the input results of sampling with bit rate.

The LCL filter can simplify equivalently under the situation that does not reduce performance as follows: footpath in the change is so that to real part operation (rather than to whole complex signal operation) and change down the footpath so that imaginary part is operated (rather than to conjugate complex number signal operation).Because to replacement LCL filter, resulting structures shown in Figure 13 28 is called the RI filter to this structure with reality-void (RI) filter, and may be interpreted as equivalent iteration embodiment on the mathematics of LCL filter.

Just as RI-FSE is better than LCL-FSE, the RI filter also provides the identical enforcement advantage that is better than the LCL filter: because Different L TI or the handled data path of FSE filter are real numbers, these two kinds of filters are to the adaptive desired calculating of the application of data and the filter calculating less than the complex data path.Especially, because the RI filter is equivalent to the LCL filter on mathematics, the estimated signal quality of output does not descend.

LCL-FRESH,LCL-FRESH-FSE,

LCL-FSE, RI-FSE, the time deformation type of LCL and RI

Though herein with 3 path LCL-FRESH, LCL-FRESH-FSE, LCL-FSE, RI-FSE, LCL and RI filter are described as fixing on a timeslice just as their filter coefficient, in fact when for example one or more signals ran into the timeslice border, (for example the LCL-PTV filter is such) that become when they can be made was with the change in the statistical property of holding received signal.Yet when in the system during to the controlling fully synchronously of the timeslice structure of a plurality of co-channel signals, forcing the timeslice border of a plurality of signals to abut against as much as possible is favourable together.Originate from the mobile radio station of in radio honeycomb communication system, operating in the same unit as a plurality of signals, then be certain to occur this required synchronous; Similarly, originate from the mobile radio station of in different units, operating, then can make required synchronous appearance by the cooperation of base station in these different units as a plurality of signals.

Adjacent channel disturbs the difference that suppresses to improve scheme

The said filter structure in front is ideally suited in suppressing co-channel interference and separating co-channel desired signal and remove distorted signals.Yet, these characteristics that the various filter structure of this class can also use adjacent channel to disturb outside the spectrum correlation properties of co-channel signal, thereby the inhibition that provides adjacent channel to disturb.When adjacent channel signals is more much better than than required channel signals (for example by closely-problem far away causes), suppress the adjacent channel interference capability with regard to particular importance.Above-mentioned filter structure can be promoted (or change) these characteristics (perhaps, if desired, only using the characteristic of adjacent channel interference) to use adjacent channel to disturb outside the cyclostationarity of required channel signals.For example, 3 path LCL-FRESH filter structures can be extended to and comprise two additional path, and every upper and lower adjacent channel that may exist is respectively with one.The signal that exists in these two additional path be height correlation from the interference that desired signal is degenerated of corresponding adjacent channel; Therefore, correctly these additional path and other three paths are made up linearly, can reduce adjacent channel significantly and disturb.Explain gained 5 path LCL-FRESH filter structure 30, wherein x among Figure 14 _-(t) and x ₊(t) be respectively under with the complex-envelope line signal (promptly the corresponding RF carrier down-conversion from them is a complex baseband) of last adjacent channel.

Perhaps, disturb as only needing to suppress adjacent channel, then the simplification as follows that 5 path LCL-FRESH filter structures 30 shown in Figure 14 can be as shown in Figure 15 to Figure 17: deletion inputs to filter g ₊And g _-Two paths, produce a new filter structure that only has three paths, its function is limited to be eliminated adjacent channel and disturbs rather than jointly eliminate adjacent channel and co-channel interference, separation signal and correcting distortion.These dual input dual paths LCL-FRESH filter structure 30a, the 30b that are shown in respectively among Figure 15 (2I-LCL-FRESH-U) and Figure 16 (2I-LCL-FRESH-L) can or use input x ₊(t) with frequency displacement 2f _c-f _b/ 2 in last adjacent channel or use input x _-(t) with frequency displacement 2f _c+ f _b/ 2 from removing interference in the adjacent channel down.As iterative scheme, can use (3I-LCL-FRESH-UL) shown in Figure 17 3 input 3 path LCL-FRESH filter 30c side by side from last and under adjacent channel in all remove interference.

According to aforementioned description of obtaining the mode of LCL-PTV, LCL-FRESH-FSE, LCL-FSE, RI-FSE, LCL and RI filter structure in 3 path FRESH filters, the people who is familiar with signal processing technology knows and these filter structures popularizations can be suppressed adjacent channel interference capability to improve their.Can change the cyclostationarity of these filter structures, promote thereby finish this class to utilize adjacent channel to disturb.

At last, notice the combination that advantageously to use different filter structure output variables.For example, each under each in two copies of 3 path LCL-FRESH filters can be used for and the last adjacent channel is estimated with the improvement that adjacent channel signals is provided; These can be estimated frequency displacement rightly, filtering adaptively then, in the received signal of required channel, deduct then, thereby reduce the adjacent channel interference.The gained signal just can be by 3 path LCL-FRESH filter process to separate co-channel signal and correcting distortion.The adjacent channel interference elimination filter or can with co-channel separate with distortion correction filter associating adaptive, perhaps adaptive separately.

Filter is concluded

Can understand, each aforementioned filter structure comprises filter (comprising time shift device and linear combiner), frequency shifter, polarization switch (ask conjugator, or real part or imaginary part selector), and adder (linear combiner).Therefore, the common core element of all these filter structures is time shift device, frequency shifter, polarization switch and linear combiner.Below table 1 conjugation period frequency and input and output sample rate are mapped as the frequency shift value and the polarization switch function (for example conjugation or real part or imaginary part are extracted) of decision specific filter structure described herein

Table 1

The conjugation period frequency, I/O sample rate	Filter structure
		±f b/2,O/O	Fig. 4, and LCL-FRESH (Fig. 5, LCL-PTV)
±f b/2,O/B	Fig. 7, LCL-FRESH-FSE-2 (Fig. 6,8,9,10,11)
		±f b/2,B/B	Figure 12, LCL (Figure 13)
2f c-f b/2,O/O	Figure 15,2I-LCL-FRESH-U
		-2f c+f b/2,O/O	Figure 16,2I-LCL-FRESH-L
?2f c-f b/2,-2f c+f b/2,O/O	Figure 17,3I-LCL-FRESH-UL
		±f b/2,-2f c+f b/2,2f c-f b/2,O/O	Figure 14,5-path LCL-FRESH

F wherein _c ⁺The carrier frequency of=following adjacent channel, f _c ^-The carrier frequency of=following adjacent channel, 0=over-extraction sample and B=bit rate sampling.Listed filter structure and the not equivalence on mathematics of the basic structure in bracket in the bracket, the former can handle by the standard block diagram and derive from the latter.

Still should understand, each aforementioned filter structure is the baseband architecture that designs for complex base band signal, these complex base band signal are corresponding to the required down-conversion channel of its center at the zero-frequency place, if and the center of required channel is at non-zero-frequency place, then can be with the standard base band to the passband transformation applications in these baseband filtering structures to obtain to be used to handle their corresponding passband filter structure of passband signal.

Joint spatial-temporal filtering and decomposition filtering

As explaining among Figure 18 and Figure 19, any one one of all can be used in the following dual mode in the above-mentioned filter structure: (1) is used for joint spatial-temporal filter structure 32, wherein each antenna output variable is by a processing in these filters, and then with the output variable addition of filtering, or (2) are used for space-time decomposition filter structure 34, the array data that pure spatial filter is used to receive wherein is then with one in these filters data of handling these space filterings.Last structure has ability and obtains than latter more performance, but cost is bigger complexity.Iteratively, 3 path LCL-FRESH filters can or substitute with the LCL-PTV filter among Figure 18 or Figure 19 or with 5 path LCL-FRESH filters, or with the LCL-FRESH-FSE among Figure 18 or Figure 19, LCL-FSE, RI-FSE, any one substitutes in LCL and the RI filter, estimates with the bit rate sampling that produces a desired signal waveform

Multiple scheme: single and a plurality of transducers,

And single and a plurality of desired signals

After using the spectrum redundancy theory of cyclo-stationary signal, we are proving on the mathematics: with the best implementation method of FRESH filter capacity and space filtering (multisensor linear combination) ability combination be make before with a plurality of data set linear combinations data from each transducer by its oneself FRESH filter, thereby reappear best single-sensor FRESH filter.The particular form of modified line conjugation space time filter during the cycle consequently.For example, use the single-sensor 3 path LCL-FRESH filters of M transducer of band can produce 3 path M transducer LCL-FRESH (3 path MS-LCL-FRESH) space time filters.Therefore, can reach a conclusion, LCL-FRESH, LCL-PTV, LCL-FRESH-FSE, LCL-FSE, RI-FSE, LCL and RI filter structure not only can be used for single-sensor receiver (wherein x (t) is that LTI or FSE filter are single input filters in invariant signal and each signal path) and also can be used for multisensor receiver (wherein x (t) is that LTI or FSE filter in vector signal and each signal path are many input filters).Similarly, LCL-FRESH, LCL-PTV, LCL-FRESH-FSE, LCL-FSE, RI-FSE, LCL and RI filter structure can be used for estimating that (wherein LTI or FSE filter are single output filters to single desired signal in each signal path; In the case, estimate a plurality of signals, then can implement a plurality of copies of filter structure concurrently as hope) also can be used for estimating a plurality of desired signals (wherein LTI or FSE filter are many output filters in each signal path).Should also be understood that in the filter used frequency shifter and ask conjugator and real part and imaginary part extractor to become multi-input device.The following fact is emphasized in these observations: the novelty of these filter structures and practicality do not lie in the structure signal dimension in any concrete signal path to a great extent, and the signal that is the pattern of conjugation, real part and imaginary part selection, frequency displacement and sampling operation and accepts these conversions forms the mode of linear combination.

The enforcement of LTI and FSE filter

The people who is familiar with signal processing technology knows that linear time invariant (LTI) filter and the FSE filter of forming the different filter structures of the present invention can implement with different modes.In this class is implemented, the LTI filter is embodied as the FIR filter, and with the overlapping of the quick Fournier transform of known use (FFT) and add or overlap and deposit algorithm and be used for its input.In another this class is implemented, use a tapped delay line, multiplier and an adder of being connected to each tap are implemented the LTI filter.In another was implemented, available infinite impulse response (IIR) filter rather than FIR filter were implemented the LTI filter; Yet, it should be noted that it may be the last costliness of calculating and/or insecure adopting this class filter.

The demodulation of restoring signal

At last, must consider reciprocation between filter structure (be used for disturb to suppress and/or co-channel signal separates) and the demodulator (being used for recovering bit stream) from the desired signal estimation.Several demodulators are used for the GMSK signal now: based on the maximum-likelihood sequence estimator (MLSE) of Viterbi algorithm, and many signals differential sensor, and decision feedback equalizer (DFE).Any one can be used for after the filter structure with interested estimated signal demodulation in these demodulators.Because DFE itself is a filter structure, therefore iterative scheme can be reasonably with the demodulation function of being implemented by conventional DFE and interference inhibition and/or the merging of Signal Separation function implemented by the spatio-temporal filtering structure, the gained filter structure will have the desired character with respect to low-down complexity for the receiver of MLSE or many signals differential detection, and, may be able to access the bit stream of the recovery of better quality because the quantity of the adjustable integral coefficient of same memory length is less.

Because the filter apparatus of all iteration described herein comprises the ability of the feedforward part of DFE, be people are just put DFE simply afterwards at filter apparatus (no matter it is single-sensor or multisensor) feedback fraction with the result of filter apparatus and DFE merging.Therefore, this feedback fraction is individually as demodulator.Be used for adjusting adaptively the algorithm of this demodulator not within the scope of the invention.

Similarly, the complexity of MLSE demodulator is lowered, and the effect of the distortion of FRESH filter structure removal ability can improve its performance.For example, by means of distortion remove or even just distortion lower, the requirement as the channel estimator of the integral part of MLSE demodulator is alleviated.D. adaptation method

Following adaptation method is discussed in the scope of gsm system, and gsm system is to use a kind of wireless communication system of the present invention.The people who is familiar with communication system and signal processing technology knows that these adaptation methods also can be used for using any other communication system of the signal with contained training sequence and norm number.

In GSM, each timeslice continues 577 μ s (156.25 bit periods during 270.833 thousand bps), and constitutes the position as shown in Figure 20.3 tail bit at main pulse string two ends all are zero, the guard time that continues 8.25 bit periods be used to prevent adjacent timeslice owing to difference in batch propagation delay bump against.116 message in each sheet can or be digitize voice (in phonetic synthesis, convolutional encoding and interleave after) or for control information.Be contained in 26 middle training sequences of sheet and be used for existing conventional GSM receiver so that for example before the MLSE demodulation, finish channel estimation.

Among the present invention, not only use this training sequence easily, and very useful normal envelope (or the norm number) characteristic of use GMSK and other msk signal is with the matched filter coefficient.Now explain four adaptation algorithm, wherein preceding two is conventional and latter two is novel exclusive with the present invention:

1. conventional adaptive equalization of training;

2. the blind dark adaptation equilibrium of norm number (CM);

3. training augmentation norm number (TACM) meropia dark adaptation equilibrium; And

4. training constraint norm number (TCCM) meropia dark adaptation equilibrium.Preceding two adaptation algorithm provide block structure for latter two algorithm.All these four algorithms are intended to seek the filter coefficient value (this matrix column number is made as and equals to wait to estimate that desired signal number and line number are made as the quantity of weight coefficient in the linear combiner that equals to be adjusted adaptively) of the matrix W that is used for controlling linear combiner.This linear combiner with these complex weighted coefficients with to comprising the element of vectorial y (t) that a plurality of different frequency displacements that postpone from the data of receiving of M transducer also may be the forms (being expressed as vector x (t)) of conjugation, to produce a signal estimation $\hat{S}\left(t\right)=W^{H}y\left(t\right)$ Vector, wherein subscript H represents conjugate transpose.For example, in 3 path LCL-FRESH filters, establish in three filters each by a tapped delay line (TDL) that comprises L tap and back with linear combiner implement.Then signal is estimated

Can be expressed as

(4) wherein y (t) defines in obvious mode.Also can be LCL-FRESH-FSE, LCL-FSE, RI-FSE, LCL, also have any expansion of these filter structures to constitute similar y (t) definition with RI filter structure and 5 path LCL-FRESH filters, number of path in these expansions, conjugation pattern or real part and imaginary part selector, the selection of frequency displacement all can change with the origin of different paths input signal.Therefore, in being described below, because therefore all filter structures of the present invention can not refer again to the specific dependency of y (t) to x (t) with this common framework representative.In addition, describe, only consider to have the situation of single-row W for simplifying; For emphasizing that this matrix deteriorates to a vector in the case, represent one group of filter coefficient with W.

The adaptive equalization of conventional training

In the adaptive equalization of routine training, receiver knows that required user is from time t ₀₁Begin to time t ₀₂Finish training signal s of emission ₀(t).This situation and the data of receiving are used to seek W.Specifically, so select W so that training period (promptly at whole t ∈ T ₀During this time, T wherein ₀=(t ₀₁, t ₀₂)) training signal s ₀(t) with estimation

Between the time average variance be minimum: $\underset{w}{\min }<\widehat{|s\left(t\right)-s_0\left(t\right){|}^2}{>}_{t\&Element;T_0}\&DoubleLeftRightArrow;w{\left(R_{\mathrm{yy}}^{\left(T_0\right)}\right)}^{-1}{R}_{{\mathrm{ys}}_0}^{\left(T_0\right)}---\left(5\right)$ Wherein Expression t ₀₁≤ t≤t ₀₂Time average during this time, and the correlation matrix between any two vectors (or scalar) a (t) and the b (t) by ${R_{\mathrm{ab}}}^{\left(t_0\right)}=<a\left(t\right)b^H\left(t\right){>}_{t\&Element;T_0}$ Determine.With reference to gsm system and Figure 21, T ₀It is the time interval that comprises contained training signal.The method is easy to implement, (for example the multichannel method of modulus or affine projection algorithm fast) computational complexity is not high when any in minimum power (RLS) algorithm or the different fast algorithm with iteration, when training sequence length surpassed filter number of coefficients (being the length of W), it was restrained reliably and apace.Yet, when having strong co-channel interference and/or when the length of W was equally matched or bigger than the latter with training signal, this conventional training adaptation equalization algorithm may not can provide enough reliable or high-quality signal to estimate.

The blind dark adaptation equilibrium of norm number (CM)

Opposite fully with conventional training adaptation equalizer, the blind dark adaptation equalizer of norm number (CM) is ignored any training signal information fully.Generation be that it uses the following fact: GMSK and other msk signal to have norm number (normal envelope) in this uses, this specific character can be made a mess of or destroy by the existence of co-channel interference and channel distortion.The CM algorithm attempts to solve following minimization problem: $\underset{w}{\min }<|\hat{s}\left(t\right)-\frac{\hat{s}\left(t\right)}{\left|\hat{s}\left(t\right)\right|}{|}^2{>}_{t\&Element;T_1}-----\left(6\right)$ T in our example wherein ₁It is one group of All Time moment in the concrete timeslice.Because this is the non-linear minimization problem that does not have the closing form solution, and lowest mean square (LMS) at random gradient descent algorithm restrain to such an extent that all too is slow, so use iteration block method (so-called minimum power CM algorithm).When the k time iteration, can separate following formula to find weighing vector W _k: $\underset{w_H}{\min }<|W_k^{H}y\left(t\right)-z_{k-1}\left(t\right){|}^2{>}_{t\&Element;T_1}---\left(7\right)$ Wherein $z_{k-1}\left(t\right)=\frac{w_{k-1}^{H}y\left(t\right)}{\left|W_{k-1}^{H}y\left(t\right)\right|}$ That is, the modulus normalization item in the variance obtains in the filter output variable during a preceding iteration.This allows to find simple linear solution: $w_k={\left(R_{\mathrm{yy}}^{\left(T_1\right)}\right)}^{-1}{R}_{{\mathrm{yz}}_{k-1}}^{\left(T_1\right)}-----\left(8\right)$ Seek modulus normalization filter output variable and conciliate these two steps of this linear equation and carry out iteratively, until convergence.The CM algorithm allows under the situation of signal that lacks in training successfully with equalizer adaptation, however known it showing insecure convergence in some cases, and when the observation cycle (as for the GSM timeslice) in short-term, its reliability may be lower than required.When with the CM algorithm application when comprising the environment of a plurality of co-channel signals, another complexity of this algorithm can appear.Developed multiple target CM algorithm, be used for adaptive many group filter weight coefficients, each is used to extract a unlike signal.Then should finish a signal sort operation with identification corresponding between the signal that extracted and the user; Be endowed different training sequences as the user and then can easily obtain this correspondence.Yet it should be noted that this training signal information can be more advantageously by their matched filter weight coefficients rather than simply with the output variable classification of blind secretly adaptive filter.The situation of two adaptive algorithm processes in addition of using will be discussed below.

Training augmentation CM (TACM) meropia dark adaptation equilibrium

Each all ignores the useful information of another use significantly in conventional training adaptation equalizer and the blind dark adaptation equalizer of CM, but TACM meropia dark adaptation equalizer is counted characteristic both uses unlike them to training signal available among the GSM and norm.The power of these methods is following idea: the performance that is produced that minimizes of the convex line combination of two cost functions is better than the independent minimized performance of any one cost function in formula (5) and the formula (7), because of general two cost functions can not minimize simultaneously.The gained cost function provides as follows: $\mathrm{\&gamma;}<|W_k^{H}y\left(t\right)-z_{k-1}\left(t\right){|}^2{>}_{t\&Element;T_1}+(1-\mathrm{\&gamma;})<|W_k^{H}y\left(t\right)-s_0\left(t\right){|}^2{>}_{t\&Element;T_0}--\left(9\right)$ In used conventional training adaptation equalizer in the gsm system, T ₀Be by the shared time interval of contained training signal (Figure 20).Use the period of time T of CM algorithm ₁Can be whole timeslice, also can be the timeslice behind the deduction training signal.No matter which kind of mode, the closing form solution is provided by following formula: $w_k=[\mathrm{\&gamma;}{R}_{\mathrm{yy}}^{\left(T_1\right)}+(1-\mathrm{\&gamma;})R_{\mathrm{yy}}^{\left(T_0\right)}{]}^{-1}[\mathrm{\&gamma;}{R}_{{\mathrm{yz}}_{k-1}}^{\left(T_1\right)}+(1-\mathrm{\&gamma;})R_{{\mathrm{ys}}_0}^{\left(T_0\right)}]\text{--}\left(10\right)$ The influence of two cost functions of convex line combiner weight coefficient γ control, also control stressing just to two different types of knowledge using in them.More approach the use that 1 γ value more lays particular emphasis on the CM characteristic, more approach the use that zero value then more lays particular emphasis on known training signal.

Notice, when initial condition is zero (being W=0) and γ=1, the special circumstances of TACM occur.In the case, use in the algorithm that TACM produced known training signal iteration k=1 and ${\mathrm{<figure-callout\; id="1"\; label="w"\; filenames="A9618046500372.png,A9618046500381.png"\; state="\{\{state\}\}">w</figure-callout>}}_1={\left({R_{\mathrm{yy}}}^{\left(T_0\right)}\right)}^{-1}{R_{{\mathrm{ys}}_0}}^{\left(T_0\right)}$ The time initialization conventional CM algorithm.

Be that norm is counted under the situation of signal (for example for GMSK, the situation of CPFSK and other msk signal),

Normalization in advance is so that be used for having unit module before the TACM algorithm at it. Not that norm is counted under the situation of signal, must insert the adaptability normaliztion constant to produce normalization adaptability TACM (AN-TACM) algorithm in the superincumbent TACM cost function: $\mathrm{\&gamma;}<|w_k^{H}y\left(t\right)-z_{k-1}\left(t\right){|}^2{>}_{t\&Element;{\mathrm{<figure-callout\; id="1"\; label="T"\; filenames="A9618046500372.png,A9618046500381.png"\; state="\{\{state\}\}">T</figure-callout>}}_1}+(1-\mathrm{\&gamma;})<|w_k^{H}y\left(t\right)-c^{*}{s}_0\left(t\right){|}^2{>}_{t\&Element;{\mathrm{<figure-callout\; id="0"\; label="T"\; filenames="A9618046500371.png,A9618046500372.png"\; state="\{\{state\}\}">T</figure-callout>}}_0}$ Formula (11) wherein C is a complex scalar.For separating this problem, can make with respect to W _k ^*And C ^*Plural gradient equate to obtain

This equation group is solved by following

Wherein $R_{11}=\mathrm{\&gamma;}{R}_{\mathrm{yy}}^{\left(T_1\right)}+(1-\mathrm{\&gamma;})R_{\mathrm{yy}}^{\left(T_0\right)},R_{12}=-(1-\mathrm{\&gamma;})R_{{\mathrm{ys}}_0}^{\left(T_0\right)},R_{21}=R_{s_{0}y}^{\left(T_0\right)},$ ${\mathrm{<figure-callout\; id="22"\; label="R"\; filenames="A9618046500451.png"\; state="\{\{state\}\}">R</figure-callout>}}_{22}=-R_{s_0{s}_0}^{\left(T_0\right)}$

Training constraint CM (TCCM) meropia dark adaptation equilibrium

TCCM and TACM difference are that it attempts to utilize two class knowledge.Specifically, when the quantity of filter coefficient during more than or equal to the training signal sample number, the used linear system of conventional training adaptation equalizer (being contained in the formula (5)) was lowly to determine.Also promptly, training signal only is enough to limit W _kIn a stator space but be not enough to it definite fully.In the case, can only say W so far _kProvide by following formula: $w_k={\left(R_{\mathrm{yy}}^{\left(T_0\right)}\right)}^{+}{R}_{{\mathrm{ys}}_0}^{\left(T_0\right)}+v_k=w^{\left(T0\right)}+v_k,$ Formula (15) is subscript () wherein ^*Expression Moore-Penrose pseudo inverse matrix, V _kBe

Any vector in the kernel, and Define in obvious mode.To V _kConstraint can guarantee at time interval T ₀In reappear training signal exactly.Use the CM characteristic to select V then _k, wherein k is expressed as in the piece minimum side of CM algorithm implements before the iteration.Be cost function after the amount of obtaining, only simply with formula (15) substitution formula (7), and with respect to V _kAsk and minimize and obtain following solution

Wherein And V ₀Row be with

But the characteristic vector of override feature value associated.

Can use iterative algorithm: at first find the solution the CM weighing vector, then this vector is projected to affine space according to formula (8) To all C _k) on.Separate at last into

And therefore at time interval T ₀Interior W _k ^HY (t) reappears training signal exactly.As desired signal ₀(t) have the norm number, as GMSK, the situation of CPFSK and other msk signal is the same, then as for TACM,

Answer normalization to obtain unit module.At last, can rely on the adaptability normalization technology of using in the derivation AN-TACM algorithm each expansion in two forms of TCCM algorithm is had the very desired signal of digital-to-analogue number to hold.These expansions can realize straight from the shoulder, therefore no longer describe in detail herein.

Additional adaptation method

Outside aforementioned TACM and TCCM method and their different expansions and special circumstances, also can use other adaptation method to come coefficient value in the selective filter.For example, under the control of some method (for example any one in the preceding method), carry out the adaptive phase I in the so-called adaptive decision direction method, estimate to produce a medium signal to high quality, and then demodulation in addition, the position of the detection of use gained is to form a known training signal in adaptive second stage, and the latter can use (for example taking advantage of method by conventional minimum) with initially improving training data.The preferred embodiment of E. overall receiver

Figure 22 has described a preferred embodiment according to overall receiver device 36 of the present invention.Down-convert to the RF signal coherence of M antenna 38 outputs complex baseband and by 40 digitlizations of relevant M antenna collecting device, its output variable is that digitlization discrete-time series M * 1 is complex vector located, all is designated as x (t) herein.This vector that receives data is promptly as the input variable of filter apparatus 42, and another input variable of this equipment 42 is the filter coefficient (W) that is provided by adaptability equipment 44, and it produces d * 1 a complex vector located signal estimative figure discrete-time series

In adaptation procedure, the switch 46 that is arranged in filter apparatus 42 outputs is delivered to adaptability equipment 44 with filter apparatus 42 output variables; When the convergence operation back (for example after the iteration of finishing pre-determined number) that detects adaptability equipment 44, switch 46 promptly moves and is last signal estimation with filter apparatus 42 output variables

Deliver to for example demodulator of next processor, the latter with the execute bit decision-making to recover binary data stream by desired signal was transmitted.Output variable (being designated as y (t) herein) also is used as the input variable of adaptability equipment 44 in the middle of one of filter apparatus 42.

Figure 23 shows the preferred embodiment of relevant M antenna collecting device 40.The customer-furnished RF tuned frequency of this equipment value and being used to is controlled the A/D sampling clock frequency values of local oscillator 48 and is controlled, and local oscillator 48 is respectively applied for real number RF is down-converted to complex baseband (being also referred to as homophase and quadrature-phase) and is used for A/D converter 50.Finishing the RF front end 52 of down-conversion can be implemented with the different modes known to the people who is familiar with the RF designing technique.The sampling clock frequency of A/D converter preferably equals bit rate f _bIntegral multiple.Remove in LCL and RI filter and need f _bSample rate outside, 2f _bRepresentative is crossed well compromise between multidata and the relevant too much amount of calculation in that temporal resolution that is provided by over-sampling and over-sampling are caused.Each digitlization discrete time homophase and orthogonal signalling to the A/D converter output combine to form and to be expressed as single complex values digitlization discrete-time signal.So each hardware chain with the individual antenna associated that forms has a signal, and it is complex vector located that such M signal is combined into digitlization discrete time M * 1 of the data of receiving.

The preferred embodiment that shows filter apparatus 42 among Figure 24.M * 1 of the filter apparatus 42 acceptance data of receiving is complex vector located and come the filter coefficient output variable of adaptivity equipment 44 as input variable.Institute's data of receiving are at first by frequency displacement, and time shift and polarization switch (be conjugation, real part is selected or imaginary part is selected) equipment 54 is handled, and its purpose is to form vectorial y (t).So the data y (t) of conversion writes down in a plurality of iteration that are stored in the adaptive algorithm of implementing herein with permission y (t) in the signal buffer memory 56 and can use for adaptability equipment 44, and signal buffer memory 56 can be implemented with dynamic RAM chips or other high speed storing medium.Matrix-vector multiplier 58 is used for the data of receiving with filter coefficient, thereby produces the output variable that can use in the next iteration (the k+1 time) that adaptability equipment 44 is finished

Show frequency displacement among Figure 25, the preferred embodiment of time shift and polarization switch equipment 54, this equipment 54 have the concrete filter structure corresponding to the 3 path LCL-FRESH filters of before having described herein.Should understand other filter of describing before also can using iteratively.M * 1 vector x (t) that this equipment is accepted the data of receiving produces the vectorial y in 3LM * 1 (t) as output as importing and connect according to formula (4).Tapped delay line (TDL) 60 produces the group of L the time shift form of a plurality of their corresponding input variables.Be to ask the polarization switch 62 of conjugator to use signal processing to be explained in this embodiment with the calibrated bolck symbol in communicating by letter with frequency shifter 64.Similarly, Figure 26 to Figure 32 shows the frequency displacement with different filter structures respectively, the different embodiment of time shift and polarization switch equipment 54, this equipment 54 has respectively corresponding to LCL-FRESH-FSE (2 iteration embodiment), LCL-FSE (2 iteration embodiment), RI-FSE (can be RI-FSE and obtain secondary iteration embodiment, be similar to LCL-FRESH-FSE and LCL-FSE), the concrete filter structure of LCL and RI filter, wherein the M signal vector of output is 4LM * 1 or 2LM * 1.Notice that the contained training sequence that occurs is used for LCL-FSE in desired signal, RI-FSE, before LCL and the used adaptation method of RI filter, these sequences must be carried out bit rate sampling and frequency displacement j by bit rate sampler 66 ^-kNotice equally, estimate as the not frequency displacement of need acquisition desired signal, then should be at filter output conjugation frequency displacement j ^kCompensate this frequency displacement j ^-kYet many GMSK demodulators require necessarily with they output variable frequency displacement j ^-k, and so conjugation frequency displacement j ^kWith frequency displacement j subsequently ^-kCancel each other out.Be also noted that, because RI-FSE and RI filter use real number M signal vector (and subsequently in application of simplifying them on the mathematics and adaptive) and do not use the plural M signal vector that uses in LCL-FSE and LCL filters, so they can more effectively be implemented.

With reference to Figure 33, with functional block diagram form of explanation polarization switch 62.Herein as can be seen, polarization switch 62 is represented gang's Design of Filter parameter, and wherein the real part 68 and the imaginary part 70 of complex input signal selected or repel by selector 72 according to period frequency of conjugation shown in the table 1 and I/O sample rate.Be also noted that if select the input signal imaginary part, then its symbol or do not change or be inverted by switch 74.In other words, complex input signal or ask conjugate perhaps selects its real part, perhaps selects its imaginary part.Can now separate, what polarization switch 62 preferably was embodied as individual feature asks conjugator or real part selector or imaginary part selector, rather than is used for the equipment that all these functions are selected on iteration ground.

With reference to Figure 34, shown the preferred embodiment of tapped delay line (TDL) 60.This is a standard implementation known in the technology.The LM that is produced * 1 output vector b (t) thus be defined as by any M * 1 input vector a (t) simply $b\left(t\right)=\begin{array}{}\left[\begin{array}{c}a\left(t\right)\\ a(t-1)\\ .\\ .\\ .\\ a(t-L+1)\end{array}\right]\end{array}$ Formula (19) therefore, through base unit sampling delay equipment and signal path are done suitably to revise, available TDL60 or hold the real number input or hold the plural number input.

Referring now to Figure 35 and Figure 36, show respectively and select TACM and TCCM as the iteration embodiment of the adaptability equipment 44 of adaptive algorithm, reach d * 1 vector ₀(t)

In all d desired signal all have the constant unit module.Adaptability equipment 44 reaches the M signal data y (t) that signal buffer memory 56 in the filter apparatus 42 is provided by 58 calculated output signal of matrix-vector multiplier in the filter apparatus 42

All be used as input.It provides the matrix W of filter factor _kAs output, it also controls the switch 46 of output shown in Figure 22 simultaneously.Adaptability equipment 44 also is used for filter apparatus 42 output variables are delivered to the low high position that is moved to of adaptability equipment 44 with switch 46 from it, and filter apparatus 42 output variables are to be suitable for d * 1 vector that the desired signal used subsequently for suitable demodulator is estimated on this high position The people who is familiar with technical application (for example programmable digital signal process chip technology) will be understood that can be directly since then in the file to directly deriving the iteration embodiment of adaptability equipment 44 in the multi-form description of TACM and TCCM.

As can be seen, the present invention can be used for extracting interested signal in numerous frequency spectrum overlapping signals of communication, separates and remove the adjacent channel interference signal of distortion and inhibition GMSK or MSK type in co-channel interference signal.These signals have real number and imaginary number component on feature, show frequency spectrum and time overlapping, demonstrate time redundancy, have twice that its value equals their carrier frequency half the conjugation period frequency of their data bit rates that adds deduct, and show the conjugation spectrum of spectral component redundant with the frequency of being separated by their conjugation period frequency.The invention provides the real part of sophisticated signal or the time shift of imaginary part, frequency displacement and conjugation or extraction are estimated to produce interested signal.Though above description comprises many features, these should not think the restriction scope of the invention, and only provide the explaination of some existing preferred embodiments of the present invention.Therefore the scope of the invention should be determined by appended claims and their legal equivalents.

Claims (12)

1. one kind is used for the equipment that extracts interested signal at the input signal that overlaps on the frequency spectrum and on the time from a plurality of, described input signal comprises the numerical data with bit rate, described input signal has carrier frequency, described input signal have equal their carrier frequency two extraordinarily and subtract half the conjugation period frequency of their data bit rates, it is redundant that described input signal shows the conjugation spectrum for the spectral component with the frequency of being separated by described conjugation period frequency, described input signal shows time redundancy, and described equipment comprises:

(a) time shift apparatus is used to produce the output signal of time shift, and wherein said interested signal is time shift;

(b) frequency displacement device is used to produce the output signal of frequency displacement, amount of being determined by its period frequency of wherein said interested signal frequency shift; And

(c) linear combination device is used for described output signal weighted sum summation to produce the estimation of described interested signal.

2. equipment described in the claim 1, wherein said input signal has real number and imaginary number component, and described equipment also comprises the polarization switch device, be used to produce a polarization output signal, the described real number of wherein said interested signal and/or imaginary number component are selected or be ostracised and wherein said real number and imaginary number component have the symbol that is changed or remained unchanged by described polarization switch device, wherein said polarization output signal is time shift, and wherein said polarization output signal is sued for peace by described linear combination device weighted sum.

3. equipment described in the claim 1 further comprises receiving system, is used to receive described a plurality of on the frequency spectrum and the input signal that overlaps on the time.

4. equipment described in the claim 1 further comprises demodulating equipment, is used for extracting data from the described estimation of described interested signal.

5. one kind is used for the equipment that extracts interested signal at the signal of communication that overlaps on the frequency spectrum and on the time from a plurality of, described signal of communication has real number and imaginary number component, described signal of communication has carrier frequency, described signal of communication comprises the numerical data with bit rate, described signal of communication shows time redundancy, described signal of communication have equal their carrier frequency two extraordinarily and subtract half the conjugation period frequency of their data bit rates, it is redundant that described signal of communication shows the conjugation spectrum for the spectral component with the frequency of being separated by described conjugation period frequency, and described equipment comprises:

(a) time shift apparatus is used to produce the output signal of a time shift, and wherein interested signal is time shift;

(b) polarization switch device is used to produce a polarization output signal, and the described real number of wherein said interested signal and/or imaginary number component are selected or be ostracised, and described real number and imaginary number component have by described polarization switch device and change or the symbol that remains unchanged;

(c) frequency displacement device is used to produce the output signal of a frequency displacement, amount of being determined by its period frequency of wherein said interested signal frequency shift; And

(d) linear combination device is used for described output signal weighted sum summation to produce the estimation of described interested signal.

6. equipment described in claim 5 also comprises receiving system, is used to receive described a plurality of on the frequency spectrum and the signal of communication that overlaps on the time.

7. equipment described in claim 6 also comprises demodulating equipment, is used for extracting data from the described estimation of described interested signal.

8. one kind is used for the equipment that extracts interested signal at the signal of communication that overlaps on the frequency spectrum and on the time from a plurality of, described signal of communication comprises the numerical data with bit rate, described signal of communication has carrier frequency, described signal of communication have equal their carrier frequency two extraordinarily and subtract half the conjugation period frequency of their data bit rates, it is redundant that described signal of communication shows the conjugation spectrum for the spectral component with the frequency of being separated by described conjugation period frequency, described signal of communication shows time redundancy, and described equipment comprises:

(a) sensing device is used to receive described signal of communication;

(b) filter is used for the contained described interested signal frequency shift of described signal of communication, and time shift and polarization are switched, amount of being determined by its period frequency of wherein said interested signal frequency shift;

(c) be used for the device that described filter is adaptive; And

(d) be used to produce the device of the estimation of described interested signal.

9. equipment described in the claim 8, wherein said sensing device comprises an antenna and a radio-frequency transmitter.

10. equipment described in the claim 9 further comprises the signal buffer storage, is used for buffer memory output signal and M signal of generation from described filter, with the input variable of the described adaptive device of opposing.

11. equipment described in the claim 9 further comprises the multiplication of vectors device that is connected to described signal buffer storage and described adaptive device, is used for signal weighting and linear combination from described filter.

12. equipment described in the claim 11 further comprises demodulating equipment, is used for extracting the contained numerical data of described estimation of described interested signal.

Images