CN102638810A - Channel modeling and simulating platform based on multidimensional channel component power spectral density - Google Patents

Abstract

The invention discloses a channel modeling and simulating platform based on multidimensional channel component power spectral density, and the channel modeling and simulating platform is novel and proposed by using an assumption that a channel consists of a plurality of diffusion components. Each component is described by virtue of delay, an arrival direction, a left direction, and Doppler frequency power spectral density. According to the channel modeling and simulating platform based on multidimensional channel component power spectral density, channel model parameters are extracted from actually-measured data by using algorithms, such as a space-alternating generalized expectation-maximization (SAGE), based on the estimation on the power spectral densities of the components. The channel modeling and simulating platform based on multidimensional channel component power spectral density can improve model accuracy and can be widely applied to the high-accuracy testing and modeling of a radio propagation channel.

Description

Channel Modeling emulation platform based on multidimensional channel component power spectral density

Technical field

The present invention relates to wireless communication field, especially a kind of new platform of setting up emulation accidental channel model can be widely used in the high precision measurement and the modeling of radio propagation channel.

Background technology

Mobile communication technology is one of hot spot technology of tool development potentiality and wide market prospects.Future mobile communications must be that capacity is bigger, speed wide-band mobile communication system higher, with better function.How improving the message capacity and the quality of GSM, is the developing focus of wideband wireless mobile communication all the time.And be to address this problem the measure of being adopted all with wireless channel necessary relation to be arranged; That is to say that the Channel Modeling problem has become research basic problem that wideband wireless mobile communication should be appreciated that, the work on this direction has important theoretical research and is worth and actual application prospect widely.The Channel Modeling emulation platform is the support platform of performance evaluation and design broadcasting system.The modeling of wireless channel always is the difficult point in the wideband wireless mobile communication system design.But should be noted that, because the complexity of mobile environment can not be set up single model.Different model never draws with concluding in the measured data of propagating environment, and certain scope of application is all arranged, and when carrying out the system engineering design, the selection of model is very important, and different sometimes models can provide different results.

Existing model comprises that the drawback of SCM model and WINNER II model is in generation model, and the path is based on " discrete minute surface path " this hypothesis and obtains.In many cases, this hypothesis has been proved to be unpractical, particularly at indoor scene.When parameter Estimation, use this hypothesis will cause producing non-existent virtual route in a large amount of reality.These virtual routes provide wrong reflection channel, and this has influenced the accuracy of model parameter conversely again.In addition, the clustering procedure that adopts in the conventional channel model depends on many didactic settings, and this causes the uncertainty of model to increase.Because the mistake in path estimation and the cluster process, the statistics of cluster can be different from truth, and therefore, the applicability of these models worse and worse.

At present, most of available channel models are to adopt method of geometry to come the propagation of artificial antenna electric wave in spatial domain, time domain and frequency domain.Model parameters such as spreading parameter are based on to be estimated to obtain to the minute surface path parameter.But when independent path Dispersion was very little, like outdoor propagation sight, using a plurality of isolated minute surface paths to describe channel was unique effective method.Therefore, for indoor, or intensive city scene, utilization minute surface path hypothesis is carried out parameter Estimation and can be produced non-existent path in a lot of reality.This an open question has reduced the applicability of model in many cases.

Summary of the invention

The objective of the invention is in order to solve many drawbacks of existing platform; Such as, traditional channel modeling method is based on " discrete minute surface path " this hypothesis and estimates channel parameter, causes producing non-existent virtual route in a large amount of reality; Influenced the accuracy of model parameter; And the clustering procedure that adopts in the conventional channel model depends on many didactic settings, causes the problems such as uncertainty increase of model, provides a kind of new platform---carry out parameter Estimation based on multidimensional channel component power spectral density; To the platform of Channel Modeling emulation, thereby improve model accuracy.The method part that the present invention is superior to the conventional channel modeling is: in new method, thereby the present invention's ability direct estimation test data obtains the characteristic of channel second moment, like this can be more accurate on the diffusion parameter of estimating channel component.In addition, the false path that produces in traditional modeling method can be suppressed, and that is to say that the channel model that utilizes the present invention to obtain more meets reality.

For reaching above purpose, the solution that the present invention adopted is:

A kind of Channel Modeling emulation platform based on multidimensional channel component power spectral density (PSD), it be utilize another kind of hypothesis promptly a channel form by a plurality of " diffusion. component " and propose a kind of new modeling and simulating platform.Each component is described by the power spectral density of delay, arrival direction, departure direction and Doppler frequency.In the present invention, the channel model parameter is based on the estimation to the power spectral density of component, utilizes space-alternating broad sense expectation maximization (SAGE) algorithm scheduling algorithm from measured data, to extract.

Further, suppose that in the present invention a channel impulse response is made up of a plurality of diffusion. component.This hypothesis is that this hypothesis that is made up of a plurality of " discrete minute surface paths " is different with the channel impulse response path that conventional model is used.Hypothesis of the present invention; Be in the channel existence of a plurality of diffusion. component be derive from experimental result such as document [1] ([1] A.Richter and R.S. " Parametric modelling and estimation of distributed diffuse scattering components of radio channels; " COST273; Tech.Rep.TD-03-198,2003.) shown in.In addition, because the finite resolving power of checkout equipment, the path with similar parameter can not be well by discrimination, and when the path is very approaching, they will be considered to be the same local expansion component of (as in time delay domain, angle domain and Doppler territory) that has.Therefore, the performance of using the characteristic of diffusion. component independently to describe a channel is rational.

Specific parameter can be described the statistical of channel in the typical environment in channel model.These statisticss can be described by the single order and the second moment of channel power spectrum density in a plurality of diffusion dimensions.The hypothesis of and incoherent scattering steady based on broad sense, the present invention one by one estimates the power spectral density of diffusion. component, combining them then obtains the power spectral density of whole channel.The feasibility of the method for this acquisition channel power spectrum density is at document [2] ([2] J.Kiefer, " Sequential minimax search for a maximum, " Proceedings of the American Mathematical Society; Vol.4, pp.502-506,1953.) and document [3] ([3] T.Betlehem; T.D.Abhayapala, and T.A.Lamahewa, " Space-time MIMOchannel modelling using angular power distributions; " In Proceedings of the 7th Australian Communications Theory Workshop; Perth, Australia, February; 1-32006 obtains proof in pp.165-170.).

Innovation of the present invention comprises:

(1) the present invention is that utilization is set up channel model with the method that power spectral density (PSD) is the basis.In the modeling process of the present invention's utilization, " cluster " this step is unnecessary.This also is a reason why can make the method for " pre-polymerization class " with PSD, and it is to be used for collecting specific parameter distribution function of use and the path with similar parameter.

Power spectral density is used for the modeling of path cluster, or on the basis based on the principle of maximum entropy under the parameter θ constraints, derives so-called spread channels component.Describing the method for the PSD of cluster path and diffusion. component confirms.

$\mathrm{\θ}=({\mathrm{\μ}}_{\mathrm{\τ}},{\mathrm{\μ}}_{\mathrm{\υ}},{\mathrm{\μ}}_{{\mathrm{\Ω}}_1},{\mathrm{\μ}}_{{\mathrm{\Ω}}_2},{\mathrm{\σ}}_{\mathrm{\τ}},{\mathrm{\σ}}_{\mathrm{\υ}},{\mathrm{\σ}}_{{\mathrm{\Ω}}_1},{\mathrm{\σ}}_{{\mathrm{\Ω}}_2},{\mathrm{\ρ}}_{{\mathrm{\Ω}}_1{\mathrm{\Ω}}_2},{\mathrm{\ρ}}_{{\mathrm{\Ω}}_1\mathrm{\τ}},{\mathrm{\ρ}}_{{\mathrm{\Ω}}_1\mathrm{\υ}},{\mathrm{\ρ}}_{{\mathrm{\Ω}}_2\mathrm{\τ}},{\mathrm{\ρ}}_{{\mathrm{\Ω}}_2\mathrm{\υ}},{\mathrm{\ρ}}_{\mathrm{\τ\υ}})---\left(1\right)$

In formula (1), K=1,2.... represent the average of ripple departure direction (DoD) and ripple arrival direction (DoA), μ _τRepresent the time delay average, μ _υRepresent the Doppler frequency average of channel power spectrum, these parameters can use following formula to calculate:

${\mathrm{\μ}}_{{\mathrm{\Ω}}_k}={\∫}_{A^{*}}{\mathrm{\Ω}}_{K}f\left(\mathrm{\ψ}\right)\mathrm{d\ψ},k=\mathrm{1,2}....---\left(2\right)$

${\mathrm{\μ}}_{\mathrm{\τ}}={\∫}_{A^{*}}\mathrm{\τf}\left(\mathrm{\ψ}\right)\mathrm{d\ψ},---\left(3\right)$

${\mathrm{\μ}}_{\mathrm{\υ}}={\∫}_{A^{*}}\mathrm{\υf}\left(\mathrm{\ψ}\right)\mathrm{d\ψ}---\left(4\right)$

This has the random vector ψ=(Ω of element ₁, Ω ₂, τ υ) represents DoD respectively, DoA, time delay and Doppler frequency.

F (ψ) is the power spectrum of propagation channel.F (ψ) is at DoD, DoA, the expansion in time delay and Doppler frequency territory σ _τ, σ _υ, can use computes:

${\mathrm{\σ}}_{{\mathrm{\Ω}}_k}=\sqrt{1-{\left|{\mathrm{\μ}}_{{\mathrm{\Ω}}_k}\right|}^2},k=\mathrm{1,2},---\left(5\right)$

(1) ρ in _(...)Can use computes

It represents the cross correlation of the power spectrum expansion of any two dimensions of appointment among the ψ.When the parameter in (1) has occurrence, be constraint below needs satisfy through the PSD that maximum entropy calculates:

Constraint 1: ${\∫}_{A^{*}}{\mathrm{\Ω}}_{1}f\left(\mathrm{\ψ}\right)\mathrm{D\ψ}---\left(12\right)$

Constraint 2: ${\∫}_{A^{*}}{\mathrm{\Ω}}_{2}f\left(\mathrm{\ψ}\right)\mathrm{D\ψ}---\left(13\right)$

Constraint 3: ${\∫}_{A^{*}}\mathrm{\τ\; f}\left(\mathrm{\ψ}\right)\mathrm{D\ψ}---\left(14\right)$

Constraint 4: ${\∫}_{A^{*}}{\mathrm{\τ}}^{2}f\left(\mathrm{\ψ}\right)\mathrm{D\ψ}---\left(15\right)$

Constraint 5:

Constraint 6: ${\&Integral;}_{A^{*}}{\mathrm{\&Omega;}}_1^{\mathrm{<figure-callout\; id="1"\; label="T\; R"\; filenames="HDA0000150012580000022.png,HDA0000150012580000023.png"\; state="\{\{state\}\}">T</figure-callout>}}{R}_{}{}_1\mathrm{\&tau;\; f}\left(\mathrm{\&psi;}\right)\mathrm{D\&psi;}---\left(17\right)$

Constraint 7: ${\&Integral;}_{A^{*}}{\mathrm{\&Omega;}}_2^T{R}_2\mathrm{\&tau;\; f}\left(\mathrm{\&psi;}\right)\mathrm{D\&psi;}---\left(18\right)$

Constraint 8: ${\&Integral;}_{A^{*}}\mathrm{\&upsi;\; f}\left(\mathrm{\&psi;}\right)\mathrm{D\&psi;}---\left(19\right)$

Constraint 9: ${\&Integral;}_{A^{*}}{\mathrm{\&upsi;}}^{2}f\left(\mathrm{\&psi;}\right)\mathrm{D\&psi;}---\left(20\right)$

Constraint 10: ${\&Integral;}_{A^{*}}{\mathrm{\&Omega;}}_1^{\mathrm{<figure-callout\; id="1"\; label="T\; R"\; filenames="HDA0000150012580000022.png,HDA0000150012580000023.png"\; state="\{\{state\}\}">T</figure-callout>}}{R}_{}{}_1\mathrm{\&upsi;\; f}\left(\mathrm{\&psi;}\right)\mathrm{D\&psi;}---\left(21\right)$

Constraint 11: ${\&Integral;}_{A^{*}}{\mathrm{\&Omega;}}_2^T{R}_2\mathrm{\&upsi;\; f}\left(\mathrm{\&psi;}\right)\mathrm{D\&psi;}---\left(22\right)$

Constraint 12: ${\&Integral;}_{A^{*}}\mathrm{\&tau;\; \&upsi;\; f}\left(\mathrm{\&psi;}\right)\mathrm{D\&psi;}---\left(23\right)$

R _(...)Represent some spin matrix. last, PSD can be write as

$f_{\mathrm{MaxEnt}}\left(\mathrm{\&psi;}\right)=\exp \{{\mathrm{<figure-callout\; id="0"\; label="b"\; filenames="HDA0000150012580000032.png,HDA0000150012580000033.png"\; state="\{\{state\}\}">b</figure-callout>}}_0+b_1^T{\mathrm{\&Omega;}}_1+b_2^T{\mathrm{\&Omega;}}_2+{\mathrm{<figure-callout\; id="3"\; label="b"\; filenames="HDA0000150012580000021.png"\; state="\{\{state\}\}">b</figure-callout>}}_3\mathrm{\&tau;}+{\mathrm{<figure-callout\; id="4"\; label="b"\; filenames="HDA0000150012580000021.png"\; state="\{\{state\}\}">b</figure-callout>}}_4{\mathrm{\&tau;}}^2+{\mathrm{<figure-callout\; id="5"\; label="b"\; filenames="HDA0000150012580000021.png"\; state="\{\{state\}\}">b</figure-callout>}}_5{\mathrm{\&Omega;}}_2^T{R}_1{R}_2^T{\mathrm{\&Omega;}}_1+{\mathrm{<figure-callout\; id="6"\; label="b"\; filenames="HDA0000150012580000021.png,HDA0000150012580000023.png"\; state="\{\{state\}\}">b</figure-callout>}}_6{\mathrm{\&Omega;}}_1^{\mathrm{<figure-callout\; id="1"\; label="T\; R"\; filenames="HDA0000150012580000022.png,HDA0000150012580000023.png"\; state="\{\{state\}\}">T</figure-callout>}}{R}_{}{}_1\mathrm{\&tau;}+b_7{\mathrm{\&Omega;}}_2^T{R}_2\mathrm{\&tau;}+{\mathrm{<figure-callout\; id="8"\; label="b"\; filenames="HDA0000150012580000023.png,HDA0000150012580000043.png"\; state="\{\{state\}\}">b</figure-callout>}}_8\mathrm{\&upsi;}+{\mathrm{<figure-callout\; id="9"\; label="b"\; filenames="HDA0000150012580000021.png,HDA0000150012580000023.png"\; state="\{\{state\}\}">b</figure-callout>}}_9{\mathrm{\&upsi;}}^2$

$\left(24\right)$

$=+{\mathrm{<figure-callout\; id="10"\; label="b"\; filenames="HDA0000150012580000022.png,HDA0000150012580000023.png"\; state="\{\{state\}\}">b</figure-callout>}}_{10}{\mathrm{\&Omega;}}_1^{\mathrm{<figure-callout\; id="1"\; label="T\; R"\; filenames="HDA0000150012580000022.png,HDA0000150012580000023.png"\; state="\{\{state\}\}">T</figure-callout>}}{R}_{}{}_1\mathrm{\&upsi;}+{\mathrm{<figure-callout\; id="11"\; label="b"\; filenames="HDA0000150012580000025.png"\; state="\{\{state\}\}">b</figure-callout>}}_{11}{\mathrm{\&Omega;}}_2^T{R}_2\mathrm{\&upsi;}+{\mathrm{<figure-callout\; id="12"\; label="b"\; filenames="HDA0000150012580000025.png,HDA0000150012580000032.png"\; state="\{\{state\}\}">b</figure-callout>}}_{12}\mathrm{\&tau;\&upsi;}\},$

b ₀Be normalization factor,

b ₁, b ₂And b ₃Be respectively to obtain b according to constraint 1,2,3 ₄-b ₁₂4-12 obtains according to constraint.The present invention representes the parameter of PSD with a vectorial b

B=(b ₁, b ₂, b ₃, b ₄, b ₅, b ₆, b ₇, b ₈, b ₉, b ₁₀, b ₁₁, b ₁₂) b ₁, b ₂, b ₆, b ₇, b ₁₀,

Always have 30 parameters among the PSD (24).

We find that these parameters are not fully independently.This is because in derivation, also have a restrictive condition not consider, such as all directions all are unit vectors.After considering this constraints, the present invention just can be reduced to 21 with 30 parameters.

(2) the used method of estimation of carrying out parameter Estimation based on power spectrum of the present invention is novel.This is feasible for the dual-polarized spectrum parameter Estimation of sextuple degree.The special feature of this algorithm is that it can guarantee that the accuracy of estimating is constant, and reduces complexity as much as possible.This method of estimation is based on space-alternating broad sense expectation maximization (SAGE) algorithm and derives out.

(3) to be expanded to 6 dimensions by the present invention suitable equally in the past common channel model.

The present invention obtains stochastic model with three steps.Be respectively measuring process, estimation steps with set up the model step.These steps can be propagated scene according to difference and one by one carry out, and are for example indoor, outdoor, and static or dynamic channel is so a set of model that makes up like this can be suitable under different situations.Below, the present invention introduces the operation of each step briefly.

Step 1: measuring process

At first, the present invention's measured channel in true environment.The measuring equipment of this channel is commonly called the channel sounding instrument, should be able to be to time domain, and the signal that delay territory (or frequency domain) and spatial domain receive is sampled.In order to extract directed spread channels from transmitter and receiver, the reflector of measuring equipment and receiver should adopt aerial array.This requires detectable signal that enough broadbands of big (for example 50-400MHZ) should be arranged, to obtain to postpone the high-resolution (for example 2.5-20ns) in territory.In addition, the time situation about becoming under, the span of observing in the measurement is big (for example 2.5-20ms) enough, so that higher resolution (for example 0.25-2s) is arranged during the Dispersion of estimation channel in Doppler frequency domain.It should be noted that in time domain the design in the observation gap of frequency domain and spatial domain should be satisfied this hypothesis of the steady incoherent scattering of broad sense (WSSUS), that is to say that channel violent variation can not take place in observation.

Notice that the test period that is used to estimate statistical static channel power spectrum satisfy following principle when setting; The channel that allows to observe is in a test sample book; Acute variation can not take place, and the channel variation that allows is at random and can not changes the statistical nature of channel.When static situation, such requirement means that usually the relative movement distance that transmits and receives between the end does not surpass several wavelength, for example 3 to 5 wavelength.The time changing environment in, this requires the displacement of scattering object in the corresponding environment also should not surpass several wavelength.What deserves to be mentioned is that the test sample book that obtains of condition continues duration and matches with the relevant duration of channel basically thus.So we can use the relevant duration of traditional channel to plan the lasting duration of power spectrum test sample book.

Step 2: estimation steps

The present invention has designed an algorithm for estimating, and handles measurement data so that estimate the power spectral density of each component with it.The parameter model that utilization of the present invention is derived by principle of maximum entropy comes the power spectral density modeling to each component as universal model.In the present invention, the present invention advises setting up with even distribution that is limited by the specific region in the three-dimensional parameter space model of power spectral density.

Parameter Estimation for power spectral density; The present invention may use different method for parameter estimation; Such as expectation maximization (EM) algorithm; Space-alternating broad sense expectation maximization (SAGE) algorithm, multiple signal classification algorithm (MUSIC) and based on parameter Estimation (ESPRIT) algorithm of rotational invariance technology.EM and SAGE algorithm are based on alternative manner.The present invention also can merge these algorithms to obtain higher precision.For example, the present invention can use multiple signal classification algorithm (MUSIC) to search the initial value that related parameter is arranged, and upgrades these parameters with EM or SAGE algorithm then.

Step 3: set up model

In this step, the statistics of model parameter is from the power spectral density estimated parameter of each component, to extract.

(1) obtains the parameter Estimation of the power spectral density of each component;

(2) experience of generation model parameter distributes;

(3) use logarithm Gauss, functions such as Laplce and probability density are confirmed probability density;

(4) confirm the final result of various propagation scenes

The present invention uses the Channel Modeling emulation platform based on multidimensional channel component power spectral density, and a particular group utilizing inseparable path to form is simultaneously come Channel Modeling, and the path in each group is described by power spectral density.So power spectral density is a better choice when describing one group of inseparable path.And the present invention uses the estimation mode of high resolution algorithm as the basis, the power spectral density parameters of the data component of estimating to measure.The result can be used for setting up model.

In sum; The present invention is a kind of Channel Modeling emulation platform based on multidimensional channel component power spectral density, utilize a kind of hypothesis promptly a channel form by a plurality of " diffusion. component " and propose high precision measurement and the modeling that a kind of new modeling and simulating platform can be widely used in radio propagation channel.

Owing to adopted such scheme, the present invention to have following characteristics:

1. utilize novel hypothesis, Channel Modeling is convenient;

2. need not clustering algorithm, channel model is more accurate;

3. the channel that has inseparable path can better be described;

4. guaranteeing to have reduced the complexity of Channel Modeling as far as possible under the accurate condition of parameter Estimation.

Description of drawings

Fig. 1 is traditional Channel Modeling process sketch map based on minute surface path hypothesis.

Fig. 2 is the Channel Modeling process sketch map based on PSD used in the present invention.

Fig. 3 is the 6 dimension PSD estimation procedure sketch mapes that the present invention uses.

The Euclidean distance variation diagram of Fig. 4 (a) expression average time delay (mean delay).

Fig. 4 (b) is the Euclidean distance variation diagram of ripple from azimuth (AoD).

Fig. 4 (c) is the Euclidean distance variation diagram of ripple from pitch angle (EoD).

Fig. 4 (d) reaches the Euclidean distance variation diagram at azimuth (AoA) for ripple.

Fig. 4 (e) reaches the Euclidean distance variation diagram of pitch angle (EoA) for ripple.

Fig. 5 is based on the channel power delay spectrum that calculates after channel power delay spectrum that original channel calculates and the reconstruct.

Fig. 6 (a) is based on direction (DoD) power spectrum that ripple that initial data calculates leaves.

Fig. 6 (b) is based on the DoD power spectrum that the data computation of estimation obtains.

Fig. 6 (c) is the dump power spectrum.(transverse axis is pitch angle (vertical angle), and the longitudinal axis is a horizontal angle.)

Fig. 7 (a) is based on direction (DoA) power spectrum that ripple that initial data calculates arrives.

Fig. 7 (b) is based on the DoA power spectrum that the data computation of estimation obtains.

Fig. 7 (c) is the dump power spectrum.(transverse axis is pitch angle (vertical angle), and the longitudinal axis is a horizontal angle.)

Embodiment

Following the present invention will specifically set forth experimental procedure and detailed details of operation, and these are for the reader understanding, or even themselves operation to obtain estimated result be very important.

The concrete steps of channel modeling method based on PSD proposed by the invention are described by Fig. 2, and as a comparison, Fig. 1 has described traditional based on the method to the modeling of minute surface path estimation.Can find out, the method that the present invention proposes at the parameter extraction block place with traditional different.Method shown in Figure 2 can be used for making up different propagation channel models, and is outdoor as indoor, static state or dynamic channel.Introduce the operation of each step execution of Fig. 2 demonstration below.

As an example, the data that the present invention handles are " 20100707PLC058by8.rbs ", because data volume is very big, in order to save the processing time, the present invention has selected representational a part of data to estimate.The data selection delay scope of handling is that the sample of [250,339] is estimated.

The 1st step, the universal model of clear and definite channel power spectrum.

The present invention needs the universal model of predefined channel power spectrum.Master copy invention according to different has different model available.In the present invention, the present invention advises using a parameter model of being derived by principle of maximum entropy.Consequent sextuple universal model still has certain novelty, because considered dual polarization in (1) modeling; (2) definition of the correlation between the polarization is novel.In addition, universal model is not limited to any specific forms.Such as, the present invention also can utilize the even distribution that is defined in a certain specific region to confirm the shape of PSD.This to a certain extent than maximum entropy be the basis pattern more directly perceived.

In the 2nd step, obtain the parameter that from the PSD of each component, estimates.

Parameter Estimation for power spectral density; The present invention may use different method for parameter estimation; Such as expectation-maximization algorithm, space-alternating broad sense expectation maximization (SAGE) algorithm, multiple signal classification algorithm (MUSIC) and based on parameter Estimation (ESPRIT) algorithm of rotational invariance technology.EM and SAGE algorithm are based on alternative manner.The EM that the present invention uses when estimating the PSD parameter with Fig. 3 explanation and the general operational process of SAGE algorithm.

(1) use the special wave beam forming method of Bart's Lay that the PSD of each component is carried out initialization.

Utilize the special wave beam forming method of Bart's Lay to calculate the DoD and ripple arrival direction (DoA) power spectrum of each time delay (delay), find the spectral function that can make the DoD-DoA-delay territory to reach maximum value.

(2) fixed delay is calculated the DoD power spectrum.

Utilize the DoD power spectrum of special wave beam forming method each time delay of calculating of Bart's Lay (delay), find the maximum of power spectrum.Through relatively these are composed, the present invention can obtain the estimated value to two channel diffusion. component then.

(3) calculate the DoA power spectrum according to known DoD

After the estimation that obtains DoD on each delay of a diffusion. component, the present invention can estimate the DoA on each delay.Utilization of the present invention is the special wave beam forming method of Bart's Lay equally.

When to delay, after the estimation of DoD and DoA was accomplished, the present invention can calculate the amplitude of the components of different polarization.The present invention brings in constant renewal in the PSD parameter with advanced iterative algorithm, reaches stable up to estimated value, exports this estimated value then.

(4) Euclidean distance between calculating parameter

In the space-alternating broad sense expectation-maximization algorithm (PSD-SAGE) based on power spectral density, the present invention has calculated the covariance matrix of sample and based on the Euclidean distance between the covariance matrix of the sample of model reconstruct.This distance can be used for the performance of check algorithm.

Such as for the parameter at PSD center of gravity place, Euclidean distance is just smoother.Like Fig. 4; The present invention can see that this is utilization PSD-SAGE first component of algorithm mean parameter time delay (mean delay) after the 5th iteration; The azimuth that ripple leaves (AoD); The azimuth (AoA) that ripple arrives, the pitch angle that ripple leaves (EoD), the Euclidean distance of the pitch angle (EoA) that ripple arrives changes.

Table 1 has provided the estimated result to each diffusion. component parameter that after 10 iteration, obtains.

Table 1

～the 5 step of the 3rd step, modelling

In these steps, the statistics of model parameter extracts from the PSD of estimative each component.The present invention sets channel model of the present invention has similar parameter setting with TR25.996 with WINNER II model.

Channel and the original channel of last the present invention after the following aspects reconstruct more once.

(1) delay power spectrum

Fig. 5 representes is based on the channel power delay spectrum that calculates after channel power delay spectrum that original channel calculates and the reconstruct.The latter all obtains through the value on the covariance matrix diagonal of each delay is made even, and these are based on, and estimated result calculates.We can see in the delay scope from Fig. 5, and the channel of reconstruct is similar with original channel.We can find out the difference of these two time delay power spectrum (PDP) equally.This mainly is because the present invention defines the scope of delay.The present invention can obtain a complete channel through increasing dynamic range.

(2) DoD, DoA power spectrum

Fig. 6 and Fig. 7 have described DoD, the DoA power spectrum that calculates based on initial data and obtain based on the calculation of parameter of estimating respectively.We see that the original spectrum of the aspect ratio of residual spectra is highly low from figure.This shows that the estimation of PSD is accurately.

Several conclusions below The above results can obtain:

1. this method that the data of selected delay scope are handled is feasible.Can reduce the complexity of calculating and channel performance is produced big influence with this method.

With based on the SAGE algorithm of minute surface different be that the PSD-SAGE algorithm can be estimated a spot of diffusion. component among the bin.The parameter reconstruct channel that utilization estimates also calculates its power spectrum, and we find that the power spectrum that it and actual measurement are come out is consistent.

The present invention uses the Channel Modeling emulation platform based on multidimensional channel component power spectral density, and a particular group utilizing inseparable path to form is simultaneously come Channel Modeling, and the path in each group is described by power spectral density.The present invention can pass through the estimation to the power spectral density of multidimensional channel component impulse response, therefrom extracts the big or small scale parameter of statistical model.

The above-mentioned description to embodiment is can understand and use the present invention for ease of the those of ordinary skill of this technical field.The personnel of skilled obviously can easily make various modifications to these embodiment, and needn't pass through performing creative labour being applied in the General Principle of this explanation among other embodiment.Therefore, the invention is not restricted to the embodiment here, those skilled in the art should be within protection scope of the present invention for improvement and modification that the present invention makes according to announcement of the present invention.

Claims (10)

1. Channel Modeling emulation platform; It is characterized in that: set a channel in this Channel Modeling emulation platform and be made up of a plurality of " diffusion. component ", each component is described by the power spectral density of delay, arrival direction, departure direction and Doppler frequency.

2. Channel Modeling emulation platform according to claim 1; It is characterized in that: use Channel Modeling emulation platform based on multidimensional channel component power spectral density; A particular group utilizing inseparable path to form is simultaneously come Channel Modeling, and the path in each group is described by power spectral density; Use the estimation mode of high resolution algorithm, the power spectral density parameters of the data component of estimating to measure as the basis.

3. Channel Modeling emulation platform according to claim 1 is characterized in that: specific parameter can be described the statistical of channel in the typical environment in channel model, and statistics is described by the single order and the second moment of channel power spectrum density in a plurality of diffusion dimensions; The hypothesis of and incoherent scattering steady based on broad sense is one by one estimated the power spectral density of diffusion. component to combine them and obtain the power spectral density of whole channel then.

4. Channel Modeling emulation platform according to claim 1 is characterized in that: its channel model parameter is based on the estimation to the power spectral density of component, utilizes algorithm from measured data, to extract;

5. Channel Modeling emulation platform according to claim 4 is characterized in that: said algorithm comprises space-alternating broad sense expectation-maximization algorithm.

6. Channel Modeling emulation platform according to claim 1 is characterized in that: the step of obtaining stochastic model comprises: measuring process, estimation steps and set up the model step; These steps are propagated scene according to difference and are one by one carried out.

7. Channel Modeling emulation platform according to claim 6 is characterized in that:

Said measuring process: at first, the present invention is measured channel in true environment; To time domain, postpone the signal that territory or frequency domain and spatial domain receive and sample; Detectable signal should have enough big broadband, to obtain to postpone the high-resolution in territory; In addition, the time situation about becoming under, the span of observing in the measurement should be enough big, so that higher resolution is arranged during the Dispersion of estimation channel in Doppler frequency domain; Should satisfy the hypothesis of the steady incoherent scattering of broad sense in the design in the observation gap of time domain, frequency domain and spatial domain;

Preferably: the broadband of 50-400MHZ; Obtain the high-resolution of the 2.5-20ns in delay territory; The span of observing in the measurement is 2.5-20ms; Resolution when in Doppler frequency domain, estimating the Dispersion of channel is 0.25-2s.

8. Channel Modeling emulation platform according to claim 6 is characterized in that:

Said estimation steps: the model of setting up power spectral density with even distribution that is limited by the specific region in the three-dimensional parameter space; For the parameter Estimation of power spectral density, the method for parameter estimation of use comprises expectation-maximization algorithm, space-alternating broad sense expectation-maximization algorithm, multiple signal classification algorithm and based on the parameter estimation algorithm of rotational invariance technology; Or,

These algorithms are merged to obtain higher precision; Preferably, search the initial value that related parameter is arranged, upgrade these parameters with expectation maximization or space-alternating broad sense expectation-maximization algorithm then with multiple signal classification algorithm.

9. Channel Modeling emulation platform according to claim 6 is characterized in that:

The said model step of setting up: the statistics of model parameter is from the power spectral density estimated parameter of each component, to extract:

(1) obtains the parameter Estimation of the power spectral density of each component;

(2) experience of generation model parameter distributes;

(3) confirm probability density;

(4) confirm the final result of various propagation scenes.

10. Channel Modeling emulation platform according to claim 9 is characterized in that: used function comprises logarithm Gauss, Laplce and probability density function when confirming probability density in the step (3).

Images