Summary of the invention
The object of the present invention is to provide a kind of broadband wireless sensing network channel estimation method of handling based on amplification forward collaboration, to be implemented in the high accuracy in the wireless sensor network with many via nodes collaboration communication, the channel estimating of low complex degree.
In order to achieve the above object, the broadband wireless sensing network channel estimation method of handling based on amplification forward collaboration provided by the invention, it comprises step: 1) in a broadband wireless sensor network, determine Chu sequence length value as the transmitting terminal of information source node, and estimate in the described broadband wireless sensor network as each relay of via node with as the noise power of the receiving terminal of information destination node and determine the maximum non-zero tap number of channel time domain response between each node as training sequence; 2) determine the power partition coefficient of described transmitting terminal and each relay according to the noise power of length value, maximum non-zero tap number and the estimation determined, so that the summation of each power partition coefficient of determining is 2, all square evaluated error minimum of the channel of while under the power partition coefficient condition of determining; 3) described transmitting terminal selects corresponding C hu sequence as time-domain training sequence according to length value of determining and power partition coefficient, and add Cyclic Prefix in described time-domain training sequence to form source signal, more described source signal is launched at first time slot; 4) unitary matrice that adopted is separately determined according to determined length value and maximum non-zero tap number in each relay, and determines separately power normalization coefficient according to the power partition coefficient of estimated noise power and described transmitting terminal; 5) after each relay receives the source signal of described transmitting terminal emission at described first time slot, its Cyclic Prefix is removed, and according to the unitary matrice that is adopted separately and determined power normalization coefficient and power partition coefficient separately, the source signal that receives is carried out linear transformation, again the signal after the linear transformation is added new Cyclic Prefix to form each repeating signal, each relay is launched the repeating signal of each self-forming respectively at second time slot; 6) in described broadband wireless sensor network, respectively in first time slot and the second time slot received signal, and merge to obtain received signal after the source signal that receives and repeating signal removed Cyclic Prefix respectively as the receiving terminal of information destination node; 7) described receiving terminal carries out the least square channel estimating to obtain corresponding each channel parameter with described received signal.
Wherein, in step 2) in, the mean square error minimum of channel estimating is and makes
Minimum, wherein, MSE is the channel estimating mean square error, σ
2Be noise variance, L is the maximum non-zero tap number of channel time domain response between each node, p
iBe the power partition coefficient of i relay, p
1Be the power partition coefficient of transmitting terminal, K is a length value, and N-1 is the via node number,
In step 3), selected Chu sequence is:
Wherein, k is the sequence number of time-domain training sequence, and l is the integer relatively prime with k.
In step 4), i the unitary matrice N that adopt the relay
iFor with n
iK * the K that is first column element ties up circular matrix, wherein n
iFor being 1 with (2L-1) * (i-2)+L+1 element, other elements are K * 1 dimensional vector of 0, and the power normalization coefficient that each relay is determined is
Preferable, in described step 2) in, the power partition coefficient of determined each relay equates that described broadband wireless sensor network can be the virtual multi-antenna system that adopts orthogonal frequency division multiplexi or single-carrier wave frequency domain equalization technology.
In sum, the broadband wireless sensing network channel estimation method of handling based on amplification forward collaboration of the present invention is at many via nodes collaboration communication scene, least square channel estimation methods based on the optimal training sequence scheme is proposed under a kind of frequency-selective channel in the AF collaboration communication, can be applicable to multinode AF scene effectively, and precision height, complexity is low, has very strong practicality.
Embodiment
See also Fig. 1, the broadband wireless sensing network channel estimation method of handling based on amplification forward collaboration of the present invention mainly may further comprise the steps:
The first step: in a broadband wireless sensor network, as the definite Chu sequence length value of the transmitting terminal of information source node as training sequence, and estimate in the described broadband wireless sensor network as each relay of via node with as the noise power of the receiving terminal of information destination node and determine the maximum non-zero tap number of channel time domain response between each node, in the present embodiment, described broadband wireless sensor network can be the virtual multi-antenna system that adopts orthogonal frequency division multiplexi or single-carrier wave frequency domain equalization technology antagonism multipath fading.In addition, a maximum non-zero tap number average of channel time domain response can be decided according to system parameter setting and applied environment between the noise power of each relay and receiving terminal and each node, and estimates that specifically implementation method all is a prior art, so do not repeat them here.
Second step: the power partition coefficient of determining described transmitting terminal and each relay according to the noise power of length value, maximum non-zero tap number and the estimation determined, so that the summation of each power partition coefficient of determining is 2, the estimation mean square error minimum of the channel of while under the power partition coefficient condition of determining.For the broadband wireless sensor network, all square evaluated error of its channel is:
Wherein, MSE is the channel estimating mean square error, σ
2Be noise variance, L is the maximum non-zero tap number of channel time domain response between each node, p
iBe the power partition coefficient of i relay, p
1Be the power partition coefficient of transmitting terminal, K is a length value, and N-1 is the via node number,
Therefore, when the power partition coefficient of determining, can adopt following method:
1, at first make the power partition coefficient of each relay equate, i.e. p
2=p
3...=p
N, wherein, N-1 is the number of relay.
2, adopt different p again
1And p
2Calculate, finding out MSE is the p of all square evaluated error minimum value of channel correspondence
1And p
2Power partition coefficient as transmitting terminal and each relay.
The 3rd step: described transmitting terminal selects corresponding C hu sequence as time-domain training sequence according to length value of determining and power partition coefficient, and add Cyclic Prefix in described time-domain training sequence to form source signal, again described source signal is launched at first time slot, in the present embodiment, the selected Chu sequence of transmitting terminal is:
Wherein, K is a length value, and k is the sequence number of time-domain training sequence, and l is the integer relatively prime with k.
The 4th step: each relay is according to determined length value and the definite unitary matrice that is adopted separately of non-zero tap number, and according to the definite power normalization coefficient separately of the power partition coefficient of estimated interchannel noise and described transmitting terminal, in the present embodiment, i unitary matrice N that adopt the relay
iFor with n
iK * the K that is first column element ties up circular matrix, wherein n
iFor being 1 with (2L-1) * (i-2)+L+1 element, other elements are K * 1 dimensional vector of 0, and the power normalization coefficient that each relay is determined is
The 5th step: after each relay receives the source signal of described transmitting terminal emission at described first time slot, its Cyclic Prefix is removed, and according to the unitary matrice that is adopted separately and determined power normalization coefficient and power partition coefficient separately, the source signal that receives is carried out linear transformation, again the signal after the linear transformation is added new Cyclic Prefix to form each repeating signal, each relay is launched the repeating signal of each self-forming respectively at second time slot, in the present embodiment, i, i ∈ (2,, N) Cyclic Prefix is removed to the signal that first time slot receives in individual relay, multiply by the unitary matrice N that it adopts
i, multiply by the power normalization factor
With power partition coefficient p
i, obtain corresponding time-domain training sequence
And then the new Cyclic Prefix of interpolation forms i repeating signal on described time-domain training sequence, and launched at second time slot.
The 6th step: in described broadband wireless sensor network, as the receiving terminal of information destination node respectively in first time slot and the second time slot received signal, and merge to obtain received signal after the source signal that receives and repeating signal removed Cyclic Prefix respectively, see also Fig. 2, information source node s by with each via node r
2... r
NCooperation is sent to information destination node d with information, and the signal that receiving terminal receives two time slots is removed respectively to merge behind the Cyclic Prefix and obtained received signal and be:
r=Sh+n
Wherein, S=[S
1S
2S
N];
()
T, ()
*, ()
H()
-1Represent vector or transpose of a matrix, conjugation, conjugate transpose and contrary respectively successively,
The expression convolution algorithm; E[] expression stochastic variable average; S
1For with
K * the L that is first column element ties up circular matrix, S
i(i=2 ..., be N) with s
RiIt is the dimension of K * (2L-1) circular matrix of first column element; L is the non-zero tap number of channel time domain response between each node; L * 1 dimensional vector h
Sd, h
Sri, h
Rid, (i=2 ..., N) the time domain tap coefficient of quasistatic multipath channel between the representation node; N represents the noise among the combined signal r.
The 7th step: receiving terminal carries out the least square channel estimating to obtain corresponding each channel parameter with described received signal, and promptly described receiving terminal utilization merging obtains received signal r and carries out the least square channel estimating:
The channel estimating of the quadratic sum minimum of error just, the channel parameter that needs in the time of can estimating channel equalization and input thus.
See also Fig. 3, it is for adopting the block diagram of the formed multinode AF of the broadband wireless sensing network channel estimation method wireless sensor network of handling based on amplification forward collaboration of the present invention, described multinode AF wireless sensor network has an information source node, a N-1 via node, reaches an information destination node, each node has only an antenna, and transceive data simultaneously, each module effect is as follows among Fig. 3:
Information source module A: the character data that generation will be transmitted.
Signal processing module B: if system adopts the OFDM technology, then B is the signal processing module that contains the IFFT computing.If system adopts the SC-FDE technology, then this module is empty.
Add CP module C: every frame data that will obtain add Cyclic Prefix.
Digital-to-analogue conversion (D/A) module D: digital signal conversion is become analog signal.
Channel module E1: the wireless multipath channel between information source node and information destination node.
Channel module E2~EN: the wireless multipath channel between an information source node and N-1 via node.
Analog-to-digital conversion (A/D) module F2~FN: analog signal is transformed into digital signal.
Remove CP module G2~GN: Cyclic Prefix is removed.
Linear process module H2~HN: via node is done the corresponding linear processing to receiving data.
Add CP module I 2~IN: every frame data that will obtain add Cyclic Prefix.
D/A module J 2~JN: digital signal conversion is become analog signal.
Wireless multipath channel between channel module K2~KN:N-1 via node and information destination node.
A/D module L: analog signal is transformed into digital signal.
Remove CP module M: Cyclic Prefix is removed.
Signal processing module N: if system adopts the OFDM technology, then N is the signal processing module that contains FFT computing and decision process.If system adopts the SC-FDE technology, then N is for containing the FFT computing, equilibrium, and the signal processing module of IFFT computing and decision process, wherein equilibrium can be selected zero forcing equalization, modes such as least mean-square error equilibrium.
Stay of two nights module O: output judgement symbol.
Below will further specify the systematic function of the communication system that adopts the broadband wireless sensing network channel estimation method of handling based on amplification forward collaboration of the present invention by emulation.The system parameters of emulation is set as:
1 information source node, 3 via nodes, 1 information destination node
Data sampling period T s is made as 4 * 10
-7s
System adopts OFDM technical antagonism multipath fading
The total carrier number K of each OFDM symbol is 128
Circulating prefix-length L is made as 16
It is that 128 Chu sequence is as time-domain training sequence that information source node is selected length
Each via node unitary matrice N
i, i ∈ (2 ..., 4) and be with n
iBe 128 * 128 dimension circular matrix, wherein n of first column element
iFor being 1 with the 31st * (i-2)+17 elements, other elements are 0 128 * 1 dimensional vectors
The power division of training sequence is: p
1=0.9042, p
2=p
3=p
4=0.3653
Frame adopts 4 yuan of quadrature amplitude modulation (4-QAM) modulation, not coding
Channel adopts COST 207 typical urban 6 footpath channel models between each node
See also Fig. 4, it is to adopt channel estimation methods of the present invention and adopt the channel estimating mean square error (MSE) of existing training sequence method at random to compare schematic diagram, and provided the MSE lower bound, training sequence is higher at random than existing to adopt precision of channel estimation that channel estimation methods of the present invention carries out as can be seen from Figure, and method of the present invention is optimum on the meaning of channel estimating mean square error minimum.See also Fig. 5 again, it is for adopting channel estimation methods of the present invention and adopting the existing error rate of system of training sequence method (BER) performance schematic diagram at random, as can be seen from the figure, adopt channel estimation methods of the present invention, system BER performance will obviously be better than training sequence method at random.
In sum, the broadband wireless sensing network channel estimation method of processing based on amplification forward collaboration of the present invention compared with prior art, it has following advantage:
1, for many relay node cooperations communication scenes, provides the method for carrying out channel estimating under a kind of frequency-selective channel in the AF wireless sensor network based on the optimal training sequence scheme.
2, adopt Chu sequence with permanent width of cloth characteristic can avoid the non-linear distortion of transmitting terminal power amplifier as training sequence, and this sequence can be produced by Direct Digital Frequency Synthesizers.
3, the design of each via node unitary matrice can make information destination node realize the least mean-square error channel estimating with quite low complexity.
4, the power division of information source node and via node can make system's minimum channel in the certain situation of general power estimate the mean square error lower bound.
Beneficial effect of the present invention:
(1) training sequence design of the present invention is simple with channel estimation methods, and highly versatile, can be used for the scene of any via node number.
(2) training sequence design of the present invention has stronger robustness with channel estimation methods, can not cause systematic function sharply to worsen along with the slight variation of channel condition.
(3) channel estimation methods of the present invention adopts least-squares algorithm, need not the channel statistical characteristic, and complexity is low.
(4) training sequence of the present invention design and channel estimation methods reduce greatly with respect to training sequence computation schemes complexity at random, and can improve the error rate of system performance effectively, and be practical, is convenient to hardware and realizes.