CN101281583A - RFID system collision resistance method based on blind signal process - Google Patents

Abstract

The invention discloses a RFID system anti-collide method based on the blind signal process, specifically discloses a method to solve the collide problem in the RFID system using the blind signal technology, namely solves the problems, provided are two or more than two labels return the information to the reader at the same tine when a plurality of labels are present in the action range of the reader, the collide problem generates, or when one label is simultaneously in the advisory area of two or more readers, two or more readers try to communicate with the label at the same tine, the collide generates. The invention markedly advances the recognition rate of the label and the recognition time; in addition the disturb of the reader and the interfere of the label are solved in the dense reader.

Description

Based on the anti-collision method of the rfid system of blind signal processing

Technical field

The invention belongs to technical field of information processing, specifically, relate to a kind of technology of utilizing blind signal to solve the collision problem in the rfid system, promptly utilize blind signal technology to solve in the reader signal reach and have a plurality of labels, synchronization has two or above label when the reader return message, the problem of collision will be produced, perhaps be in the interrogation zone of two or more readers simultaneously when a label, the same time produces the problem of collision when having two or more readers to attempt with this label communication.

Background technology

The collision of RFID (Radio Frequency Identification System) system, be meant that the RFID label is in identifying: in the reader signal reach, have a plurality of labels, synchronization has two or above label when the reader return message, to produce collision, be called the label collision; On the other hand, when a label is in the interrogation zone of two or more readers simultaneously, the same time also will produce collision when having two or more readers to attempt with this label communication, be called the reader collision, the algorithm that solves collision is called anti-collision algorithm.Many to the label reverse collision Study on Problems in the research of these two kinds of anti-collision problems of RFID.

In the near-field communication technology, RFID mainly adopts SDMA, FDMA, CDMA and four kinds of multichannel cut-in methods of TDMA.Consider the factors such as complicacy of rfid system communication form, power consumption, system, three kinds of situations of front are because the complexity that realizes, difficulty is all bigger, can only under the special situation of minority, use, therefore anti-collision research mainly concentrates on the TDMA field, and this also is to realize the anti-collision of RFID the most general method at present.

From present research situation, though the anti-collision algorithm of RFID is many, also do not reach the requirement of application fully, the space also improves a lot.

Domestic and international many label reverse collisions technology mainly concentrates on ALOHA class and the tree type class algorithm, the purpose of research mainly is round accuracy of identification that how to improve many labels and recognition efficiency, on the time complexity and communication complexity that reduces system simultaneously and consumed, this also is the target that following many label reverse collisions algorithm is asked most.The anti-collision of label wherein also has more outstanding algorithm no matter be abroad or domestic many algorithms arranged all.

Generally speaking, although the anti-collision algorithm of label is many, also comparative maturity.On the contrary, with regard to the anti-collision algorithm research of reader, all fewer both at home and abroad, all the more so especially at home.Wherein the Colorwave algorithm is the most typical in the anti-collision of reader, and in the at present fewer anti-collision algorithm of reader, the Colorwave algorithm is also quoted or introduced maximum.The anti-collision algorithm of reader that real at home own research and design is come out is quite few, and the document that quite a few is arranged is mostly just introduced the Colorwave algorithm when referring to that reader is counter to be collided.

Therefore, the anti-collision algorithm of reader is still waiting to drop into more, more deep research.

Summary of the invention

In order to solve RFID collision problem (comprising reader collision and label collision), the present invention reaches the purpose that addresses the above problem by adopting the method for blind signal processing.

Based on the anti-collision method of the rfid system of blind signal processing, it comprises based on the label reverse collision of blind signal processing with based on the reader of blind signal processing is counter collides, wherein,

Label reverse collision step based on blind signal processing comprises:

1) sets up label collision model: input signal S={s based on blind signal processing _i(t) } (i=1,2 ..., m), output signal X={x _i(t)) } (i=1,2 ..., n) (n 〉=m, n represent the number of output signal or observation signal, and m represents input signal, also are unknown signalings, wait to ask the number of signal, the t express time, and X or X (t) represent by a plurality of signal x _i(t) signal vector of Zu Chenging, x _i(t) be i signal among the signal vector X (t)), signal model is X=AS;

2) mixed signal (output signal) X is carried out the Fourier conversion, makes mixed signal become sparse signal: with signal x (t) (t=1,2 ..., T) be standardized as x ' (t) (t=1,2 ..., T);

3) to x ' (t) (t=1,2 ..., T) carry out based on Euclidean apart from cluster: with x ' (t) (t=1,2 ..., T) be divided into n class, the n that correspondence an is divided into class is designated as x successively ⁱ(t) (t=1,2 ..., T ⁱ, i=1,2 ..., n) the corresponding n of a difference source;

4) to x ⁱ(t) (t=1,2 ..., T ⁱ, i=1,2 ..., n) adopt principal component analysis (PCA) accurately to estimate the aliasing matrix A respectively, obtain the direction vector of n source signal, be designated as A=(a successively ¹, a ²..., a ⁿ);

5) calculate all kinds of x respectively ⁱ(t) (t=1,2 ..., T ⁱ, i=1,2 ..., straight line intensity Q (x n) ⁱ(t)), if Q is (x ⁱ(t)) all bigger, mean that then cluster is successful, the aliasing matrix A is estimated accurately to forward step 6) to, on the contrary repeating step 2)-5);

6) adopt linear programming to find the solution following optimization problem:

$\underset{s\left(t\right)}{\min }\underset{i}{\mathrm{\Σ}}\left|s_i\left(t\right)\right|$

s.t.As(t)＝x(t)(t＝1，2，...，T)

Symbol description in the formula: min represents content in [] is minimized, and A is the aliasing matrix, and s (t) is source signal or signal vector to be asked, and x (t) is observation or known signal vector, s _i(t) be i signal in s (t) vector, s.t. represents constraint;

7), isolate n sparse source signal, s (t)=(s at domain of variation by linear programming ₁(1), s ₂(2) ..., s _n(t)) ^T(t=1,2 ..., T);

8) to the source signal s (t) in the domain of variation thus carry out the reconstruct that contrary Fourier conversion realizes source signal accordingly, the sequence number of each label that obtains separating is realized the identification of label.

Comprise based on the anti-step of colliding of the reader of blind signal processing:

1) sets up reader collision model: input signal S={s based on blind signal processing _i(t) } (i=1,2 ..., m), near reader undesired signal U={u _i(t) } (i=1,2 ..., m), output signal X={x _i(t) } (i=1,2 ..., n) (n 〉=m, n represent the number of output signal or observation signal, and m represents input signal, also are unknown signalings, wait to ask the number of signal, the t express time, and X or X (t) represent by a plurality of signal x _i(t) signal vector of Zu Chenging, x _i(t) be i signal among the signal vector X (t)), signal model is X=AS+U;

2) variances sigma of calculating interference noise U _u

3) mixed signal (output signal) X is carried out the Fourier conversion, makes mixed signal become sparse signal: with signal x (t) (t=1,2 ..., T) be standardized as x ' (t) (t=1,2 ..., T);

4) to x ' (t) (t=1,2 ..., T) carry out based on Euclidean apart from cluster: with x ' (t) (t=1,2 ..., T) be divided into n class, the n that correspondence an is divided into class is designated as x successively ⁱ(t) (t=1,2 ..., T ⁱ, i=1,2 ..., n) the corresponding n of a difference source;

5) to x ⁱ(t) (t=1,2 ..., T ⁱ, i=1,2 ..., n) adopt principal component analysis (PCA) accurately to estimate the aliasing matrix A respectively, obtain the direction vector of n source signal, be designated as A=(a successively ¹, a ²..., a ⁿ);

6) calculate all kinds of x respectively ⁱ(t) (t=1,2 ..., T ⁱ, i=1,2 ..., straight line intensity Q (x n) ⁱ(t)), if Q is (x ⁱ(t)) all bigger, mean that then cluster is successful, the aliasing matrix A is estimated accurately to forward step 7) to, on the contrary repeating step 3)-6);

7) find the solution following optimization problem:

$\underset{s\left(t\right)}{\min }{\left[\frac{1}{{\mathrm{\σ}}_u^2}\left|\right|\mathrm{As}\left(t\right)-x\left(t\right)\left|\right|+\underset{i}{\mathrm{\Σ}}\left|s_i\left(t\right)\right|\right]}_{(t=\mathrm{1,2},...,T)}$

Symbol description in the formula: min represents content in [] is minimized, and A is the aliasing matrix, and s (t) is source signal or signal vector to be asked, and x (t) is observation or known signal vector, s _i(t) be i signal in s (t) vector;

8), isolate n sparse source signal, s (t)=(s at domain of variation by optimized Algorithm ₁(1), s ₂(2) ..., s _n(t)) ^T(t=1,2 ..., T);

9) to the source signal s (t) in the domain of variation thus carry out the reconstruct that contrary Fourier conversion realizes source signal accordingly, the sequence number of each label that obtains separating is realized the identification of label.

Described label reverse collision step 2 based on blind signal processing) can also carry out wavelet transformation to mixed signal (output signal) X in, make mixed signal become sparse signal, the corresponding inverse wavelet transform that adopts in step 8) replaces contrary Fourier conversion.

Beneficial effect of the present invention: utilize the collision problem in the rfid system of blind signal processing, the discrimination of label obviously improves, and Shi Bie time also can obviously improve simultaneously; In the otherwise address intensive reader, the interference of reader and the problems such as interference of label.

Description of drawings

Fig. 1 is reader collision synoptic diagram;

Fig. 2 is the mimo system model based on blind signal processing;

Fig. 3 is the label collision model based on mimo system;

Fig. 4 is the reader collision model based on mimo system.

Embodiment

Below in conjunction with accompanying drawing the present invention is further set forth.

On the collision problem that solves reader, all be the collision problem that solves the RFID data transmission from protocol layer or modulating layer basically, so also can consider to use signal processing technology to solve its collision problem, this also is the method for the collision of the solution reader that will take of this patent.This patent mainly is to adopt blind signal processing technology to solve the problem of reader collision.

Briefly, separate in (BSS) X=AS (output signal X={x in the formula in blind source _i} ^T, mixing constant A={a _iI=1,2 ... .m, input signal S={s _jJ=1,2 ..., do not know A and S among the n, m 〉=n).But it is independent of each other needing the hypothesis source signal, and has only a gaussian signal in the source signal at the most, utilizes some technology finally to isolate source signal S.After beginning to set up this cover theory the nineties, blind signal processing has developed into more complicated constraint condition situation still less at present.

The collision problem of reader as shown in Figure 1.When a plurality of readers read a label simultaneously, will produce collision, at this moment label is received the data X that reader sends over _i, but and do not know simultaneously owing to the conflict of reader when data transmission, not know the data S that reader sends by channel A yet _iSo, constituted MIMO (majority the go into many output) model of a blind signal processing as shown in Figure 2.Here only be a kind of situation of listing the reader collision, same, other situations also are to set up the corresponding communication model in this way.

From top simple discussion, the reader collision problem of rfid system can use the blind signal processing method to solve as can be seen.

The anti-collision method of a kind of rfid system based on blind signal processing, it comprises based on the label reverse collision of blind signal processing with based on the reader of blind signal processing is counter collides.

Label reverse collision step based on blind signal processing comprises:

1) sets up label collision model (shown in Figure 3): input signal S={s based on blind signal processing _i(t) } (i=1,2 ..., m), output signal X={x _i(t) } (i=1,2 ..., n) (n 〉=m, n represent the number of output signal or observation signal, and m represents input signal, also are unknown signalings, wait to ask the number of signal, the t express time, and X or X (t) represent by a plurality of signal x _i(t) signal vector of Zu Chenging, x _i(t) be i signal among the signal vector X (t)), signal model is X=AS;

2) mixed signal (output signal) X is carried out the Fourier conversion, makes mixed signal become sparse signal: with signal x (t) (t=1,2 ..., T) be standardized as x ' (t) (t=1,2 ..., T);

3) to x ' (t) (t=1,2 ..., T) carry out based on Euclidean apart from cluster: with x ' (t) (t=1,2 ..., T) be divided into n class, the n that correspondence an is divided into class is designated as x successively ⁱ(t) (t=1,2 ..., T ⁱ, i=1,2 ..., n) the corresponding n of a difference source;

4) to x ⁱ(t) (t=1,2 ..., T ⁱ, i=1,2 ..., n) adopt principal component analysis (PCA) accurately to estimate the aliasing matrix A respectively, obtain the direction vector of n source signal, be designated as A=(a successively ¹, a ²..., a ⁿ);

5) calculate all kinds of x respectively ⁱ(t) (t=1,2 ..., T ⁱ, i=1,2 ..., straight line intensity Q (x n) ⁱ(t)), if Q is (x ⁱ(t)) all bigger, mean that then cluster is successful, the aliasing matrix A is estimated accurately to forward step 6) to, on the contrary repeating step 2)-5);

6) adopt linear programming to find the solution following optimization problem:

$\underset{s\left(t\right)}{\min }\underset{i}{\mathrm{\Σ}}\left|s_i\left(t\right)\right|$

s.t.As(t)＝x(t)(t＝1，2，...，T)

Symbol description in the formula: min represents content in [] is minimized, and A is the aliasing matrix, and s (t) is source signal or signal vector to be asked, and x (t) is observation or known signal vector, s _i(t) be i signal in s (t) vector, s.t. represents constraint;

7), isolate n sparse source signal, s (t)=(s at domain of variation by linear programming ₁(1), s ₂(2) ..., s _n(t)) ^T(t=1,2 ..., T);

8) to the source signal s (t) in the domain of variation thus carry out the reconstruct that contrary Fourier conversion realizes source signal accordingly, the sequence number of each label that obtains separating is realized the identification of label.

Wherein, symbol description in the above-mentioned steps: vector x (t) (t=1,2 ..., T) be the mixed signal of the acceptance of reader, here x (t) (t=1,2 ..., T) be exactly the mixed sequence of a plurality of tag serial number; Vector x ' (t) (t=1,2 ..., T) be by x (t) (t=1,2 ..., T) be normalized into the normalized signal on " semi-simple circle of position "; The Euclidean distance definition: $d(x,y)=\sqrt{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{(x_k-y_k)}^2}$ X in the formula, y are respectively n * 1 dimension variable; Straight line intensity Q (x ⁱ(t)) definition: suppose signal vector Z (t)=(z ₁(t), z ₂(t) ...., z _m(t)) ^T(t=1,2 ..., T) can be expressed as T point of m-dimensional space vector, make m * m sample covariance matrix ∑ of Z (t), m eigenwert of note ∑ is λ ₁＞λ ₂＞...＞λ _m〉=0, the straight line intensity of Z (t) is defined as

$Q\left(z\left(t\right)\right)=1-(\sqrt{{\mathrm{\λ}}_2}/\sqrt{{\mathrm{\λ}}_1}).$

Comprise based on the anti-step of colliding of the reader of blind signal processing:

1) sets up reader collision model (shown in Figure 4): input signal S={s based on blind signal processing _i(t) } (i=1,2 ..., m), near reader undesired signal U={u _i(t) } (i=1,2 ..., m), output signal X={x _i(t) } (i=1,2 ..., n) (n 〉=m, n represent the number of output signal or observation signal, and m represents input signal, also are unknown signalings, wait to ask the number of signal, the t express time, and X or X (t) represent by a plurality of signal x _i(t) signal vector of Zu Chenging, x _i(t) be i signal among the signal vector X (t)), signal model is X=AS+U;

2) variances sigma of calculating interference noise U _u

3) mixed signal (output signal) X is carried out the Fourier conversion, makes mixed signal become sparse signal: with signal x (t) (t=1,2 ..., T) be standardized as x ' (t) (t=1,2 ..., T);

4) to x ' (t) (t=1,2 ..., T) carry out based on Euclidean apart from cluster: with x ' (t) (t=1,2 ..., T) be divided into n class, the n that correspondence an is divided into class is designated as x successively ⁱ(t) (t=1,2 ..., T ⁱ, i=1,2 ..., n) the corresponding n of a difference source;

5) to x ⁱ(t) (t=1,2 ..., T ⁱ, i=1,2 ..., n) adopt principal component analysis (PCA) accurately to estimate the aliasing matrix A respectively, obtain the direction vector of n source signal, be designated as A=(a successively ¹, a ²..., a ⁿ);

6) calculate all kinds of x respectively ⁱ(t) (t=1,2 ..., T ⁱ, i=1,2 ..., straight line intensity Q (x n) ⁱ(t)), if Q is (x ⁱ(t)) all bigger, mean that then cluster is successful, the aliasing matrix A is estimated accurately to forward step 7) to, on the contrary repeating step 3)-6);

7) find the solution following optimization problem:

$\underset{s\left(t\right)}{\min }{\left[\frac{1}{{\mathrm{\σ}}_u^2}\left|\right|\mathrm{As}\left(t\right)-x\left(t\right)\left|\right|+\underset{i}{\mathrm{\Σ}}\left|s_i\left(t\right)\right|\right]}_{(t=\mathrm{1,2},...,T)}$

Symbol description in the formula: min represents content in [] is minimized, and A is the aliasing matrix, and s (t) is source signal or signal vector to be asked, and x (t) is observation or known signal vector, s _i(t) be i signal in s (t) vector;

8), isolate n sparse source signal, s (t)=(s at domain of variation by optimized Algorithm ₁(1), s ₂(2) ..., s _n(t)) ^T(t=1,2 ..., T);

9) to the source signal s (t) in the domain of variation thus carry out the reconstruct that contrary Fourier conversion realizes source signal accordingly, the sequence number of each label that obtains separating is realized the identification of label.

Wherein, the algorithm of symbol description such as label in the above-mentioned steps.

Claims (2)

1, the anti-collision method of a kind of rfid system based on blind signal processing is characterized in that, its step comprises based on the label reverse collision of blind signal processing with based on the reader of blind signal processing is counter collides, wherein,

Label reverse collision step based on blind signal processing comprises

1) sets up label collision model: input signal S={s based on blind signal processing _i(t) } (i=1,2 ..., m), output signal X={x _i(t) } (i=1,2 ..., n) (n 〉=m, n represent the number of output signal or observation signal, and m represents input signal, also are unknown signalings, wait to ask the number of signal, the t express time, and X or X (t) represent by a plurality of signal x _i(t) signal vector of Zu Chenging, x _i(t) be i signal among the signal vector X (t)), signal model is X=AS;

2) mixed signal (output signal) X is carried out the Fourier conversion, makes mixed signal become sparse signal: with signal x (t) (t=1,2 ..., T) be standardized as x ' (t) (t=1,2 ..., T);

3) to x ' (t) (t=1,2 ..., T) carry out based on Euclidean apart from cluster: with x ' (t) (t=1,2 ..., T) be divided into n class, the n that correspondence an is divided into class is designated as x successively ⁱ(t) (t=1,2 ..., T ⁱ, i=1,2 ..., n) the corresponding n of a difference source;

4) to x ⁱ(t) (t=1,2 ..., T ⁱ, i=1,2 ..., n) adopt principal component analysis (PCA) accurately to estimate the aliasing matrix A respectively, obtain the direction vector of n source signal, be designated as A=(a successively ¹, a ²..., a ⁿ);

5) calculate all kinds of x respectively ⁱ(t) (t=1,2 ..., T ⁱ, i=1,2 ..., straight line intensity Q (x n) ⁱ(t)), if Q is (x ⁱ(t)) all bigger, mean that then cluster is successful, the aliasing matrix A is estimated accurately to forward step 6) to, on the contrary repeating step 2)-5);

6) adopt linear programming to find the solution following optimization problem:

$\underset{s\left(t\right)}{\min }\underset{i}{\mathrm{\Σ}}\left|s_i\left(t\right)\right|$

s.t.?As(t)＝x(t)(t＝1，2，...，T)

Symbol description in the formula: min represents content in [] is minimized, and A is the aliasing matrix, and s (t) is source signal or signal vector to be asked, and x (t) is observation or known signal vector, s _i(t) be i signal in s (t) vector, s.t. represents constraint;

7), isolate n sparse source signal, s (t)=(s at domain of variation by linear programming ₁(1), s ₂(2) ..., s _n(t)) ^T(t=1,2 ..., T);

8) to the source signal s (t) in the domain of variation thus carry out the reconstruct that contrary Fourier conversion realizes source signal accordingly, the sequence number of each label that obtains separating is realized the identification of label.

Comprise based on the anti-step of colliding of the reader of blind signal processing:

1) sets up reader collision model: input signal S={s based on blind signal processing _i(t) } (i=1,2 ..., m), near reader undesired signal U={u _i(t) } (i=1,2 ..., m), output signal X={x _i(t) } (i=1,2 ..., n) (n 〉=m, n represent the number of output signal or observation signal, and m represents input signal, also are unknown signalings, wait to ask the number of signal, the t express time, and X or X (t) represent by a plurality of signal x _i(t) signal vector of Zu Chenging, x _i(t) be i signal among the signal vector X (t)), signal model is X=AS+U;

2) variances sigma of calculating interference noise U _u

3) mixed signal (output signal) X is carried out the Fourier conversion, makes mixed signal become sparse signal: with signal x (t) (t=1,2 ..., T) be standardized as x ' (t) (t=1,2 ..., T);

4) to x ' (t) (t=1,2 ..., T) carry out based on Euclidean apart from cluster: with x ' (t) (t=1,2 ..., T) be divided into n class, the n that correspondence an is divided into class is designated as x successively ⁱ(t) (t=1,2 ..., T ⁱ, i=1,2 ..., n) the corresponding n of a difference source;

5) to x ⁱ(t) (t=1,2 ..., T ⁱ, i=1,2 ..., n) adopt principal component analysis (PCA) accurately to estimate the aliasing matrix A respectively, obtain the direction vector of n source signal, be designated as A=(a successively ¹, a ²..., a ⁿ);

6) calculate all kinds of x respectively ⁱ(t) (t=1,2 ..., T ⁱ, i=1,2 ..., straight line intensity Q (x n) ⁱ(t)), if Q is (x ⁱ(t)) all bigger, mean that then cluster is successful, the aliasing matrix A is estimated accurately to forward step 7) to, on the contrary repeating step 3)-6);

7) find the solution following optimization problem:

$\underset{s\left(t\right)}{\min }{\left[\frac{1}{{\mathrm{\σ}}_u^2}\left|\right|\mathrm{As}\left(t\right)-x\left(t\right)\left|\right|+\underset{i}{\mathrm{\Σ}}\left|s_i\left(t\right)\right|\right]}_{(t=\mathrm{1,2},...,T)}$

Symbol description in the formula: min represents content in [] is minimized, and A is the aliasing matrix, and s (t) is source signal or signal vector to be asked, and x (t) is observation or known signal vector, s _i(t) be i signal in s (t) vector;

8), isolate n sparse source signal, s (t)=(s at domain of variation by optimized Algorithm ₁(1), s ₂(2) ..., s _n(t)) ^T(t=1,2 ..., T);

9) to the source signal s (t) in the domain of variation thus carry out the reconstruct that contrary Fourier conversion realizes source signal accordingly, the sequence number of each label that obtains separating is realized the identification of label.

2, the anti-collision method of the rfid system based on blind signal processing according to claim 1, it is characterized in that, described label reverse collision step 2) can also carry out wavelet transformation to mixed signal (output signal) X in based on blind signal processing, make mixed signal become sparse signal, the corresponding inverse wavelet transform that adopts in step 8) replaces contrary Fourier conversion.

Images