CN1901521A - United state maximum likelihood difference detecting method - Google Patents

Abstract

This invention discloses a largest likelihood difference test method of a joint state, which first of all constitutes an input/output joint state variable, then computes the measurement accumulation values of the paths and the measurement of each input/output joint state and selects and preserves the paths accumulated in a same input/output joint state to be recursive according to symbols to realize the simplification of lattice patterns step by step then to judge the test result based on the patterns, which realizes the multi-symbol combined maximum likelihood test to the complexity of the test to difference modulation sequence not influenced by the variance of the sequence length.

Description

United state maximum likelihood difference detecting method

Technical field

The present invention relates to a kind of signal detecting method that is applied to Communication and Information Systems, relate in particular to a kind of united state maximum likelihood difference detecting method.

Background technology

Differential Detection need not produce coherent reference signal and channel estimating at receiving terminal as a kind of noncoherent detection scheme in the communication system, thereby has avoided that carrier phase is difficult to have the restriction of efficient recovery and has simplified the receiver design in the coherent detection.Make it become very competitive a kind of detection scheme, in communication system, obtain increasingly extensive application.

Though Differential Detection has been avoided many requirements of coherent detection, has paid the cost of performance loss.In order to remedy the detection performance loss, scholars have proposed various schemes.Be mainly many symbol classes Differential Detection algorithm (MSDD).It mainly is divided into based on the Differential Detection algorithm (MLDD) of maximum likelihood with based on the Differential Detection algorithm (DFDD) of decision-feedback.The former separates the maximum likelihood in the individual symbol of continuous L (L＞2) as testing result, the method (VDD) that it mainly can be divided into the MLDD method of detection of packets again and realize MLDD with the Viterbi algorithm.The corresponding with it VDD algorithm (RSVDD) that has occurred respectively subtracting the grouping MLDD algorithm of complexity and subtracting state waits the improvement algorithm.

Above-mentioned various many symbol detection methods, standard MLDD and VDD scheduling algorithm detection complexity increase with the sequence length exponentially that detects.Various shortcut calculations such as RSVDD are the reduction that cost exchanges complexity for to sacrifice the detection performance, and its complexity still increases with L with linear or other relation.

Summary of the invention

The objective of the invention is to solve the problems of the technologies described above, provide a kind of the maintenance to try to achieve the united state maximum likelihood difference detecting method that the random length differential of sequence detects under the constant prerequisite of low detection complexity.

United state maximum likelihood difference detecting method of the present invention comprises following steps:

The first step: structure input and output united state variable, and to its metric initialization;

Second step: that calculates each path measures measuring of accumulated value and each input and output united state;

The 3rd step: each path that converges at same input and output united state is selected to keep;

The 4th step: carry out step 2 and three by the symbol recurrence, realize the abbreviation step by step of grid chart;

The 5th step: judgement obtains testing result on the basis of grid chart abbreviation.

Compared with prior art, the present invention has following advantage:

Detection complexity is not subjected to the sequence length variable effect, guarantees that testing result is under many symbols joint-detection maximum likelihood prerequisite of separating, and has overcome existing various many symbols difference detecting methods rapid defective of rising of detection complexity when symbolic number increases.

Detecting the traditional relatively incoherent difference detecting method of performance is significantly increased.The theoretical limit that can approach many symbols maximum likelihood Differential Detection arbitrarily with the sequence length increase is promptly carried out the theoretical value of coherent detection to differential modulation (coding) signal.The present invention has been simultaneously also for the sequencal estimation of various Discrete Finite state Markov processes under the memoryless noise conditions provides a kind of new departure, its in the application of every field with increasingly extensive.

Description of drawings

Fig. 1 is the FB(flow block) of united state maximum likelihood difference detecting method of the present invention.

Fig. 2 is a M-DPSK signal grid chart.

Fig. 3 is the explanation of M-DPSK combined signal status quantitavie recursive calculation and grid chart recurrence abbreviation.

Fig. 4 is a 2DPSK signal grid chart.

Fig. 5 is the explanation of 2DPSK combined signal status quantitavie recursive calculation and grid chart recurrence abbreviation.

Embodiment

United state maximum likelihood difference detecting method of the present invention comprises following steps:

The first step: structure input and output united state variable, and to its metric initialization;

Second step: that calculates each path measures measuring of accumulated value and each input and output united state;

The 3rd step: each path that converges at same input and output united state is selected to keep;

The 4th step: carry out step 2 and three by the symbol recurrence, realize the abbreviation step by step of grid chart;

The 5th step: judgement obtains testing result on the basis of grid chart abbreviation.

Above-mentioned input and output united state variable is by the common decision of the input and output of modulator, each input and output of modulator constitute a kind of united state, the united state number is the number of combinations of the two value of input and output.

The accumulated value of measuring in each path is measured the sum of measuring with the united state of its succession for this path is current in the above-mentioned current sign.

The accumulated value of measuring of measuring each path by relatively converging at this state of above-mentioned input and output united state obtains, with path metric accumulated value the measuring as this united state of minimum.

Above-mentioned each path that converges at same input and output united state is selected to keep is to keep to have minimum and measure of accumulated value in to each path that converges at same united state, gives up the branch road that remaining can not become optimal path.

Above-mentioned input and output united state variable measure the distance of each point on the correspondence reception value of initial value by calculating follow-up first symbol in reference position and its transmitting terminal planisphere and first data estimator with it on corresponding transmitting terminal planisphere each point apart from acquisition.

The above-mentioned decision method that obtains testing result of adjudicating on the basis of grid chart abbreviation can be that piece detects, and also can be by symbol detection.

If M phase phase shift keying (MPSK) sequence is x after the energy normalized, the transmission sequence after the x process differential modulation (coding) is s, sends symbol s for any time k _kHave following form,

s _k＝x _ks _k-1 (1)

X wherein _k∈ { e ^{2 π mj/M}| k=1,2, M=0,1 ..., M-1}, M are the chopping phase figure place, s ₀For the initial reference signal, can be in differential modulation (coding) sequence more arbitrarily.

The received signal r that receiving terminal is corresponding with s can be expressed as:

r＝s+n (2)

Wherein n is memoryless noise.

According to formula (1) structure M-DPSK signal grid chart, as shown in Figure 2.Represent incoming symbol x with node _k, change corresponding output symbol s with the representative of the transfer path between adjacent node input _kFor simplified model is without loss of generality, with initial reference signal s ₀Be normalized to s ₀=1.

The first step, structure input and output united state.

For the M-DPSK signal,, will be output as s for any k symbol _k=e ^{2 π α j/M}, α=0,1 ..., M-1 is input as x simultaneously _k=e ^{2 π β j/M}, β=0,1 ..., the modulator of M-1 (encoder) state is defined as the input and output united state, is designated as φ _{K, α, β}, establishing its measurement value is η _{K, α, β}

Second step: that calculates each path measures measuring of accumulated value and each input and output united state.

The initial value η that united state is measured _{1, u, v}By calculating f (d (r ₁, s ₁), d (r ₁/ r ₀, x ₁)) obtain, wherein (α is with α β) to f, and β is the function of independent variable, and (α is α β) to d, and distance function between the β comprises various forms of distances such as Hamming distance, Euclidean distance.Such as calculating the initial value η that united state is measured with formula (3) _{1, u, v}

η _1，u，v＝d _1，u+d _1，v (3)

Wherein $\left\{\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="d"\; filenames="A20061008833000131.png,A20061008833000132.png"\; state="\{\{state\}\}">d</figure-callout>}}_{1,u}={|r_1-{\mathrm{<figure-callout\; id="2"\; label="e"\; filenames="A20061008833000132.png"\; state="\{\{state\}\}">e</figure-callout>}}^{2\mathrm{\&pi;uj}/M}|}^2\\ {\mathrm{<figure-callout\; id="1"\; label="d"\; filenames="A20061008833000131.png,A20061008833000132.png"\; state="\{\{state\}\}">d</figure-callout>}}_{1,v}={|\frac{r_1}{r_0}-{\mathrm{<figure-callout\; id="2"\; label="e"\; filenames="A20061008833000132.png"\; state="\{\{state\}\}">e</figure-callout>}}^{2\mathrm{\&pi;vj}/M}|}^2\end{array}\right.;$ u，v＝0，1，…，M-1。

The explanation of M-DPSK combined signal status quantitavie recursive calculation and grid chart recurrence abbreviation as shown in Figure 3.In the recurrence stage, according to formula (1) in grid chart, x _K-1M kind value in any one has all converged representative s _K-1The M paths of different values.While x _K-1And x _kM is arranged ²Kind of compound mode is so have M in every symbol when k＞2 ³Paths.According to s _kAnd x _kThe combination of different values, total M ²Plant united state, and each united state has all converged the M paths.

Every paths measure the measure sum of accumulated value for the united state of the distance function value between its current output valve and the received signal and its succession.Measure the comparison of accumulated value to converging at each paths in the same united state, this minimum is measured accumulated value give this united state, measure as it, for next symbol recursive calculation, as the formula (4).

η _k，u，v＝min{η _k-1，m，n+d _k，u|mod(m+v，M)＝mod(u，M)} (4)

M wherein, n, u, v=0,1 ..., M-1; K＞1; (α is β) for being that mould rems to α with β for mod; d _{K, u}Be to be output as e in k the symbol ^{2 π uj/M}The measuring of path, output valve by calculating this path corresponding modulating device (encoder) and the distance between the received signal obtain, distance can be tried to achieve by various distance functions, as Euclidean distance d commonly used _{K, u}=| r _k-e ^{2 π uj/M}| ², Hamming distance etc.

When program realizes, for any one united state φ in k the symbol _{K, u, v}, according to u, the v value is tried to achieve m, travels through all M kind values of n then, and relatively η is selected to keep in the back _{K-1, m, n}+ d _{K, u}Minimum value, and give η with it _{K, u, v}

The 3rd step: each path that converges at same input and output united state is selected to keep.

To converging at each path of same united state, keep and have minimum and measure of accumulated value, give up all the other M-1 bars.Retain the M paths in every symbol.

The 4th step: carry out step 2 and three by the symbol recurrence, realize the abbreviation step by step of grid chart.

As shown in Figure 3, the calculating that each path metric and united state are measured in the current sign in the grid chart and the selection in path and give up are carried out on the basis that measuring of last each united state of symbol tried to achieve.In each united state, recursively carry out the reservation in path and give up the abbreviation of then having realized whole grid chart by symbol.

The 5th step: judgement output on the basis of grid chart abbreviation.

To differential modulation (coding) sequence that receives, carry out aforementioned grid chart abbreviation by symbol, can guarantee that then the number of path that keeps in every symbol does not increase.Adjudicate on this basis to detect and mainly contain following two class methods: 1, piece detects, and it is measured by the united state in this moment relatively constantly in judgement, and the selection reservation has the united state that minimum is measured, and with the retention path of its correspondence as testing result.Also promptly relatively retain the accumulated value of measuring in path, get measure the accumulated value minimum one as court verdict.2, by symbol detection, it sets fixing decoding depth, when receiving a new symbol, measuring of current each united state compared, find out and have united state and the corresponding retention sequence thereof that minimum is measured, date back to the current sign distance and be the position of decoding depth, with its respective signal as testing result, when decoding depth is set long enough and can be guaranteed that the retention path of each united state can be merged into one, can directly date back to current sign position distance for adjudicating the acquisition testing result in the position of decoding depth.

Use the increase of symbolic number can improve the performance that many symbols are united Differential Detection, reduce overhead.But because detection complexity can rise significantly because of the increase of symbolic number, many symbols joint-detection of no performance loss is difficult to Project Realization.It is constant that the united state detection method that the present invention proposes can be kept complexity, tries to achieve the maximum likelihood of sequence simultaneously and separate in random length.Approach the implementation method that a practicality is provided fully for many symbols Differential Detection performance theoretical limit.While also provides a kind of new departure for the sequencal estimation of Discrete Finite state Markov process under the memoryless noise conditions.

Embodiment:

With the 2DPSK signal is that example is illustrated the present invention, and Fig. 4 is a 2DPSK signal grid chart.

The first step, structure input and output united state.

For detecting length is the 2DPSK sequence of L, according to s in k-1 the symbol _K-1And x _K-14 kinds of various combination tectonic syntaxis state φ of value _{K-1, i, j}I, j=0,1, as the formula (5), and in Fig. 5 and respectively, indicate with line style 1 to 4.

Second step: that calculates each path measures measuring of accumulated value and each input and output united state.

United state initial value η _{1, u, v}U, v=0,1 obtains by following formula.

η _1，u，v＝d _1，u+d _1，v (6)

Wherein

$\left\{\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="d"\; filenames="A20061008833000131.png,A20061008833000132.png"\; state="\{\{state\}\}">d</figure-callout>}}_{1,u}={|r_1-e^{\mathrm{\&pi;uj}}|}^2\\ {\mathrm{<figure-callout\; id="1"\; label="d"\; filenames="A20061008833000131.png,A20061008833000132.png"\; state="\{\{state\}\}">d</figure-callout>}}_{1,v}={|\frac{r_1}{r_0}-e^{\mathrm{\&pi;vj}}|}^2\end{array}\right.$

Fig. 5 is the recursive calculation of 2DPSK combined signal status quantitavie and the explanation of grid chart recurrence abbreviation, wherein k＞2.In k symbol, continue to indicate extended on its basis path with these 4 kinds of line styles.Every paths has all been inherited certain state φ _{K-1, i, j}I, j=0,1 measurement value η _{K-1, i}, j:i, j=0,1.From grid chart, point to x as can be seen _k Same value 1 or-1 all have 4 paths, two representative output s wherein _kArticle=1, two, s is exported in representative _k=-1.Wherein satisfy state φ _{K, 0,0}Be line style 1 and line style 3, satisfy state φ _{K, 1,0}Be line style 2 and line style 4, satisfy state φ _{K, 0,1}Be line style 2 and line style 4, satisfy state φ _{K, 1,1}Be line style 1 and line style 3.

The path metric accumulated value is the status quantitavie value and the current sum of measuring of a last symbol of this path succession.Current measuring by calculating receiving symbol r _kOutput s with this path _kBetween Euclidean distance obtain.Can try to achieve by following formula for measuring of each united state of 2DPSK signal shown in Figure 4:

$\left\{\begin{array}{c}{\mathrm{\&eta;}}_{k,\mathrm{0,0}}=\min ({\mathrm{\&eta;}}_{k-\mathrm{1,0,0}}+d_0,{\mathrm{\&eta;}}_{k-\mathrm{1,0,1}}+d_0)\\ {\mathrm{\&eta;}}_{k,\mathrm{1,0}}=\min ({\mathrm{\&eta;}}_{k-\mathrm{1,1,0}}+{\mathrm{<figure-callout\; id="1"\; label="d"\; filenames="A20061008833000131.png,A20061008833000132.png"\; state="\{\{state\}\}">d</figure-callout>}}_1,{\mathrm{\&eta;}}_{k-\mathrm{1,1,1}}+d_1)\\ {\mathrm{\&eta;}}_{k,\mathrm{0,1}}=\min ({\mathrm{\&eta;}}_{k-\mathrm{1,1,0}}+d_0,{\mathrm{\&eta;}}_{k-\mathrm{1,1,1}}+d_0)\\ {\mathrm{\&eta;}}_{k,\mathrm{1,1}}=\min ({\mathrm{\&eta;}}_{k-\mathrm{1,0,0}}+{\mathrm{<figure-callout\; id="1"\; label="d"\; filenames="A20061008833000131.png,A20061008833000132.png"\; state="\{\{state\}\}">d</figure-callout>}}_1,{\mathrm{\&eta;}}_{k-\mathrm{1,0,1}}+d_1)\end{array}\right.---\left(7\right)$

Wherein

$\left\{\begin{array}{c}{d}_{k,0}={|r_k-1|}^2\\ d_{k,1}={|r_k-(-1)|}^2\end{array}\right.;$ k＝2，…，L

The 3rd step: each path that converges at same input and output united state is selected to keep.

At each φ _{K, i, j}I, j=0, that carries out two paths in 1 respectively measures accumulated value relatively, keeps to have the less path of measuring accumulated value, gives up another.Keeping under the constant prerequisite of Maximum Likelihood Detection performance, only needing to keep measuring of four paths and four kinds of united states in every symbol.

The 4th step: carry out step 2 and three, the abbreviation step by step of Recursive Implementation grid chart by the symbol recurrence.

As shown in Figure 5, in the grid chart, the calculating that each path metric and united state are measured in the current sign and the selection in path and give up are carried out on the basis that measuring of last each united state of symbol tried to achieve.In each united state, recursively carry out the reservation in path and give up the abbreviation of then having realized whole grid chart by symbol.

The 5th step: judgement output on the basis of grid chart abbreviation.

After L symbol received, by measuring comparison, one that selects the cumulative path metric minimum as court verdict.Or to establish decoding depth be 0.5L, when receiving a new symbol, measuring of current each united state compared, and finds out to have the retention sequence that minimum is measured.Date back to current sign distance and be the position of decoding depth, with its respective signal as testing result.

Claims (7)

1, a kind of united state maximum likelihood difference detecting method is characterized in that, comprises following steps:

The first step: structure input and output united state variable, and to its metric initialization;

Second step: that calculates each path measures measuring of accumulated value and each input and output united state;

The 3rd step: each path that converges at same input and output united state is selected to keep;

The 4th step: carry out step 2 and three by the symbol recurrence, realize the abbreviation step by step of grid chart;

The 5th step: judgement obtains testing result on the basis of grid chart abbreviation.

2, united state maximum likelihood difference detecting method according to claim 1 is characterized in that,

Above-mentioned input and output united state variable is by the common decision of the input and output of modulator, each input and output of modulator constitute a kind of united state, the united state number is the number of combinations of the two value of input and output.

3, united state maximum likelihood difference detecting method according to claim 1 is characterized in that,

The accumulated value of measuring in each path is measured the sum of measuring with the united state of its succession for this path is current in the above-mentioned current sign.

4, united state maximum likelihood difference detecting method according to claim 1 is characterized in that,

The accumulated value of measuring of measuring each path by relatively converging at this state of above-mentioned input and output united state obtains, with path metric accumulated value the measuring as this united state of minimum.

5, united state maximum likelihood difference detecting method according to claim 1 is characterized in that,

Above-mentioned each path that converges at same input and output united state being selected to keep, is to have minimum and measure of accumulated value keeping in each path that converges at same united state, gives up the branch road that remaining can not become optimal path.

6, united state maximum likelihood difference detecting method according to claim 1 is characterized in that,

On the reception value of the initial value that above-mentioned input and output united state variable is measured by calculating follow-up first symbol correspondence in reference position and its transmitting terminal planisphere the distance of each point and first data estimator with it on corresponding transmitting terminal planisphere each point apart from acquisition.

7, united state maximum likelihood difference detecting method according to claim 1 is characterized in that,

The above-mentioned decision method that obtains testing result of adjudicating on the basis of grid chart abbreviation can be that piece detects, and also can be by symbol detection.

Images