CN1100395C - sequence estimation method - Google Patents

sequence estimation method Download PDF

Info

Publication number
CN1100395C
CN1100395C CN97194349A CN97194349A CN1100395C CN 1100395 C CN1100395 C CN 1100395C CN 97194349 A CN97194349 A CN 97194349A CN 97194349 A CN97194349 A CN 97194349A CN 1100395 C CN1100395 C CN 1100395C
Authority
CN
China
Prior art keywords
path
mentioned
state
constantly
metric
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired - Fee Related
Application number
CN97194349A
Other languages
Chinese (zh)
Other versions
CN1217837A (en
Inventor
久保博嗣
村上圭司
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Mitsubishi Electric Corp
Original Assignee
Mitsubishi Electric Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Mitsubishi Electric Corp filed Critical Mitsubishi Electric Corp
Priority to CN97194349A priority Critical patent/CN1100395C/en
Priority to AU73213/98A priority patent/AU705414B2/en
Priority to PCT/JP1997/000652 priority patent/WO1998039848A1/en
Priority to CA002253395A priority patent/CA2253395C/en
Priority to EP97903655A priority patent/EP0902545A4/en
Priority claimed from PCT/JP1997/000652 external-priority patent/WO1998039848A1/en
Publication of CN1217837A publication Critical patent/CN1217837A/en
Application granted granted Critical
Publication of CN1100395C publication Critical patent/CN1100395C/en
Anticipated expiration legal-status Critical
Expired - Fee Related legal-status Critical Current

Links

Images

Classifications

    • H—ELECTRICITY
    • H03—ELECTRONIC CIRCUITRY
    • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
    • H03M13/65—Purpose and implementation aspects
    • H03M13/6502—Reduction of hardware complexity or efficient processing
    • H—ELECTRICITY
    • H03—ELECTRONIC CIRCUITRY
    • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
    • H03M13/37—Decoding methods or techniques, not specific to the particular type of coding provided for in groups H03M13/03 - H03M13/35
    • H03M13/39—Sequence estimation, i.e. using statistical methods for the reconstruction of the original codes
    • H03M13/3961—Arrangements of methods for branch or transition metric calculation
    • H—ELECTRICITY
    • H03—ELECTRONIC CIRCUITRY
    • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M13/00—Coding, decoding or code conversion, for error detection or error correction; Coding theory basic assumptions; Coding bounds; Error probability evaluation methods; Channel models; Simulation or testing of codes
    • H03M13/37—Decoding methods or techniques, not specific to the particular type of coding provided for in groups H03M13/03 - H03M13/35
    • H03M13/39—Sequence estimation, i.e. using statistical methods for the reconstruction of the original codes
    • H03M13/41—Sequence estimation, i.e. using statistical methods for the reconstruction of the original codes using the Viterbi algorithm or Viterbi processors
    • H03M13/4107—Sequence estimation, i.e. using statistical methods for the reconstruction of the original codes using the Viterbi algorithm or Viterbi processors implementing add, compare, select [ACS] operations
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L1/00—Arrangements for detecting or preventing errors in the information received
    • H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
    • H04L1/0045—Arrangements at the receiver end
    • H04L1/0054—Maximum-likelihood or sequential decoding, e.g. Viterbi, Fano, ZJ algorithms

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Theoretical Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Error Detection And Correction (AREA)

Abstract

The sequence estimation method using the existing viterbi algorithm is modified as follows. First, a revised route of a route past at a specific time point at which the route was revised is derived. Calculating the path metric of the changed path. Comparing the path metric of the normal path with the path metric of the changed path, and correcting the path of the path when the path metric of the changed path is small. Selecting a remaining path from the plurality of paths passing through the corrected path.

Description

Serial estimating method
Technical field
The present invention relates to being in the digital data transfer of representative, according to the characteristic of received signal and transmission line, at the serial estimating method of receiver side estimation received signal sequence with automobile telephone etc.
Background technology
The technical background that explanation now the present invention relates to.
Existing technology 1.
Usually, when transmission of digital data,, can not directly receive transmission signal, and will carry out receiving after the conversion according to the state of transmission line, noise etc. from transmitter side at receiver side owing to reasons such as the state of transmission line or noises.
Figure 16 illustrates the pattern conversion of the signal on the transmission line.As shown in figure 16, signal produces and postpones on transmission line, has also added noise simultaneously.Therefore, suppose that sending signal is X k, the then following expression of received signal. r k = Σ i = 0 L c i x k - i + w k Formula (1)
In the formula, L represents to make the memory length of the transmission line that sends signal generation delay, c iThe expression tap coefficient, w kThe expression noise component(s).Tap coefficient, noise component(s) all are the values by the characteristic decision of transmission line.
Become and in receiver, receive received signal r k, and according to this received signal r kWith tap coefficient c i, estimation sends signal.
In receiver (sequence estimation device), send the candidate of signal and known tap coefficient, following estimated value (to call duplicate in the following text) of calculating received signal by overlapping input. r ^ k = Σ i = 0 L c i x ^ k - i Formula (2)
Receiver (sequence estimation device) is calculated actual received signal with following formula (3) and the error power of the estimated value (duplicate) of the received signal calculated by formula (2). Σ k | e k | 2 = Σ k | r k - r ^ k | 2 Formula (3)
Receiver (sequence estimation device) is explored the candidate that the error power that is obtained by formula (3) is the transmission signal of minimum, and this candidate is estimated as the transmission signal.
Sequence estimation when specifically, the memory length that transmission line is described is L=2 is handled.The pattern of optimal sequence estimating device when figure 17 illustrates the memory length L=2 of transmission line.The sequence estimation device can make the pattern reproduction same with the pattern of transmission line.But in sequence estimation device shown in Figure 17, in the pattern of transmission line, do not constitute the attachment device that additional noise is used.
Specifically, the sequence estimation device is by constituting with the lower part: have the memory that sends the estimated value of signal with the input of the same memory length of the memory length of transmission line; The tap coefficient of regulation is taken multiplier from the estimated value of the transmission signal of memory output; By will by multiplier take advantage of value take out addition, calculate the adder of the estimated value (duplicate) of received signal; Get from the difference counter of the difference of the received signal of the estimated value (duplicate) of the received signal of adder output and reality; And get from the quadratic power of the value of difference counter output and quadratic power and calculator.In addition, the tap coefficient of the regulation of setting in the multiplier is identical with the tap coefficient that obtains according to the characteristic of transmission line.
Utilize above sequence estimation device, the method for carrying out the best-estimated is described.
At first, obtain the candidate of the transmission signal of transmission sequence length N.
Then, this is sent in the memory of candidate's list entries estimating device of signal, multiplier is with tap coefficient c 1, c 2Take from each signal of memory output, tap coefficient c 0Take not on the signal by memory output.
Adder by will by multiplier take advantage of the whole additions of value, obtain the estimated value (duplicate) of received signal.
The difference of the estimated value (duplicate) of the received signal that in difference counter, obtains the received signal of actual reception and obtain by adder.
Then, quadratic power and calculator carry out quadratic power and calculating to the difference value from difference counter output.Quadratic power and calculator be by to quadratic power and the addition of whole bursts with the difference of the estimated value (duplicate) of received signal and received signal, obtain quadratic power and.
Supposing to send sequence length is N, then has the candidate of Nth power this transmission signal of 2, carry out above-mentioned processing to whole candidates.
The quadratic power that best determining device will be obtained by quadratic power and calculator and be estimated as the transmission signal for the candidate of hour transmission signal.
Existing technology 2.
Under above-mentioned this best situation about judging, amount of calculation increases with the amount that transmission sequence length N should take advantage of with being directly proportional.Therefore, adopted and imported a kind of best determining method that is called the method for viterbi algorithm.At G.D.Forney, the Jr. works " The Viterbialgorithm ", Proc.IEEE, Vol.61, No.3 understands viterbi algorithm in detail among the pp.268-278 (March 1973).
Under the situation of transmission line pattern shown in Figure 180,, then can calculate the error power of k constantly if be divided into transmission data and opposite with it the transmission data before 2 constantly that reach of present moment k.
As shown in figure 19, in the best of using viterbi algorithm is judged, use according to the figure (hereinafter referred to as lattice (ト レ リ ス) figure) that represents the transinformation of data through 2 data combination constantly.
This lattice Figure 19 considers following characteristic and links figure through 2 data combination constantly with line.
This characteristic is that the state of the signal of for example supposing that certain is stored in memory constantly is 00, transfers to state 10 or 00 any among both constantly at next, but does not transfer to state 01 or 11.Constantly 1 transferring under the situation of 000 shift register, cause by obtaining 000 or 100.
Therefore, when linking through 2 data combination constantly, at line between state 00 and the state 10 and between state 00 and state 00 with line.At line not between state 00 and the state 01 and between state 00 and state 11.
Make the lattice figure that has considered transfer characteristic like this.In Figure 19, so-called line is that expression can be shifted, and can not represent to cause transfer by line.In addition, the line that below will represent state transitions is called branch.In addition, in lattice Figure 19, put down in writing solid line and dotted line, solid line is represented input signal 0 state transitions, and dotted line is represented can input signal 1 state transitions.
Shown in the lattice figure among Figure 19,, then can determine to utilize it can obtain error power through 3 data combination constantly if connect through 2 data combination constantly with line.
Secondly, describe the contents processing of the viterbi algorithm that uses lattice figure shown in Figure 19 in detail.
At first, how to determine status number with viterbi algorithm, suppose that the transmission line memory length is L, then status number is 2 L power.In other words, the exponentiation of status number and transmission line memory length L increases pro rata.So operand increases along with the increase of this status number.
, in conventional example 1, must explore the candidate who all sends signal, different therewith, by using viterbi algorithm, can reduce and handle number.
Figure 20 represents that viterbi algorithm is in each processing sequence constantly.
In the explanation afterwards, the state during with each moment k * * remember and make s[k, * *], will moment k 1The time state be * *, at moment k 2Transfer to the path note of state 00 and make s[k 1, * *]/s[k 2, 0 0].
(1) at first, calculate the quadratic power error of each branch (line segment among Figure 19).
The quadratic power error of this each branch is called branch metric.
For example, in connection status s[0,00] and state s[1, under the situation of branch 00], because the data in this branching representation process moment 3 are 000, so these data be multiply by tap coefficient respectively, obtain the difference with actual received signal, calculate, calculate the quadratic power error by quadratic power.
Utilize such method, calculate the quadratic power error of whole branches.
(2) secondly, take out the path that expression reaches certain state (being 00,10,01,11 among Figure 19) state transitions before constantly.Then, path metric is calculated in the addition of the branch metric of the branch by will constituting the path of being taken out accumulative total.In addition, the calculating of path metric will be carried out whole paths of whole states.
For example, as reaching s[2,00] the path, have path s[0,00]/s[1,00]/s[2,00] and path s[0,11]/s[1,01]/s[2,00] two paths.Then, this two paths is calculated path metric.
(3) compare between the path metric of the mulitpath that each state is taken out.All states are carried out this comparison.
(4) Bi Jiao result, with path metric be minimum path as the most correct path, store this path and path metric.In addition, each state is carried out the storage of this path and path metric.
Result relatively, path metric is called remaining path metric for minimum path is called remaining path (giving birth to the residual リ バ of I ス) with the path metric in this remaining path.
For example, reaching state s[2,00] the path in, to path s[0,00]/s[1,00]/s[2,00] path metric and path s[0,11]/s[1,01]/s[2,00] path metric compare, the path that value is little becomes remaining path.
(5) in viterbi algorithm, from the mulitpath that reaches certain state, select the remaining path of the last item.
The algorithm that carries out above processing at each constantly is exactly a viterbi algorithm.
Figure 21 illustrates the result who carries out viterbi algorithm with lattice figure shown in Figure 19.Figure 21 represents the last remaining path that obtains.
Then, from the remaining path of last moment of a frame partly having been carried out above-mentioned processing, select path metric for minimum path as final path.In Figure 21, the path of representing with solid line and thick dashed line is a final path.
To be estimated as the transmission signal from the train of signal that final path obtains.
The best determining method of the viterbi algorithm that illustrated more than adopting is at G.D.Forney, Jr. " the Maximum-likelihood sequnence estimation of digitalsequences in the presence of intersymbol interference " of works, IEEETrans.Inform.Theory, Vol.IT-18, the method that is called Maximum-Likelihood Sequnence Estimation (MLSE: optimal sequence is estimated) of explanation among the No.3, pp.363-378 (May 1972).
In this MLSE, suppose that the transmission line memory length is L, then the status number of viterbi algorithm is 2 L power.Like this, according to branch's monodrome of expression state transitions insert the value of transmission line memory method be MLSE.Transmission line pattern when figure 12 illustrates L=5.When MLSE was applied to this pattern, status number was 25 powers, promptly 32.
Existing technology 3.
Above-mentioned MLSE compares with existing technology 1., can reduce and handle number, but status number is pressed the exponentiation increase of transmission line memory length L, so treating capacity is still very big.
As the method that addresses this problem is works such as A.Duel-Hallen " Delayeddecision-feedback sequence estimation ", IEEE Trans.Commun., vol.COM-37,5, pp.428-436, the method that is called Decision-Feedback Sequence Estimation (DFSE) described in the May 1989.
This method that is called DFSE is with the method after the processing change part of above-mentioned MLSE.
The difference of the work of DFSE and MLSE is described simply with Figure 15.Because the transmission line memory among Figure 15 is 5, so in order to insert total candidate, need make state with 5 memories.At this moment under the situation of MLSE, status number is 32.
On the other hand, under the situation of DFSE, though the transmission line memory is 5, the memory of using as the state of making is conceived to two memories., when making state with two memories, in order to insert the memory of transmission line, half the data deficiencies of 3 symbolic components of back.Therefore, utilize the remaining path that interrelates with the state of k-1 constantly, the value that will obtain from remaining path is used as half the data of 3 symbolic components of back.
By using such DFSE, can make status number reduce to 4 from 32.
Secondly, specifically describe the processing method of DFSE according to Figure 20.Show to Figure 20 vague generalization each contents processing constantly of DFSE.
＜the processing of k=1～3 constantly 〉
At first, suppose that the data when k is negative are known, states all when supposing k=0 all exist, and carry out following explanation.
From lattice Figure 19 as can be known, in order to reach s[1,00], s[0 is arranged, 00]/s[1,00] or s[0,01]/s[1,00] the path.Calculate their path metric, to comparing between this path metric.Then, select the little path of path metric as remaining path.Here, select s[0,00]/s[1,00] the path as remaining path.
Equally, determine to reach s[1 singly, 10], s[1,01], s[1,11] the remaining path of each state.
By carrying out same processing, also moment k=2, k=3 are determined to reach the remaining path of each state.
Processing through such before moment k=3, obtains remaining path shown in Figure 2.In addition, the path metric in the remaining path before the storage k=3.
＜the processing of k=4 constantly 〉
(the computing of branch metric: BMG-I)
Owing to become state s[4,00], thus can be state s[3,00] or state s[3,01].Therefore, as becoming state s[4,00] path of usefulness, have s[0,01]/s[1,10]/s[2,01]/s[3,00]/s[4,00] path and s[0,01]/s[1,10]/s[2,11]/s[3,01]/s[4,00] such two paths in path.
At moment k=4, from this path s[0,01]/s[1,10]/s[2,01]/s[3,00]/and s[4,00] obtain the candidate of the transmission sequence of " 000101 ", at moment k=4, from this path s[0,01]/s[1,10]/and s[2,11]/s[3,01]/s[4,00] obtain the candidate of the transmission sequence of " 001101 ".According to this value, the branch metric when calculating k=4.
(the computing of path metric: ADD-I)
Then, calculate reach state s[4,00] the path metric in above-mentioned path.
Branch metric during k=4 by will newly calculating is added in the path metric before the k=3 that calculates and storing over, obtains path metric.That is, in path s[0,01]/s[1,10]/and s[2,01]/s[3,00]/[4,00] under the situation, branch metric is added to state s[3,00] path metric in, in path s[0,01]/s[1,10]/s[2,11]/s[3,01]/situation of [4,00] under, branch metric is added to state s[3,01] path metric in.
To state s[4,10], state s[4,01], state s[4,11] carry out same processing.
(the comparison process of path metric: CMP-I)
Secondly, to reaching the state s[4 that calculates when the ADDI, 00] the path metric of two paths between compare.That is, path s[0 relatively, 01]/s[1,10]/s[2,01]/s[3,00]/path metric and the path s[0 of [4,00], 01]/s[1,10]/s[2,11]/s[3,01]/path metric of [4,00].By getting poor (that measures is poor) between the selected path metric, carry out the comparison of path metric.
(select to handle: SEL)
Then, in path s[0,01]/s[1,10]/s[2,01]/s[3,00] path metric/[4,00] and path s[0,01]/s[1,10]/s[2,11]/s[3,01] in the path metric/[4,00], select the little path of path metric as reaching s[4,00] remaining path.
To each state s[4,10], state s[4,01], state s[4,11] also carry out the processing of above CMP-I and SEL.
Carry out such processing in each moment that k=5 is later.Carry out the above-mentioned processing of frame part, compare between the path metric in the remaining path that each state of last moment is obtained, select the little path of path metric.The path of this selection is a final path.
Take out transmission sequence, the transmission sequence that the transmission sequence that estimation is obtained sends as transmitter side from this final path.
Figure 22 is the block diagram that carries out the sequence estimation device of viterbi algorithm shown in Figure 20.
In Figure 22,1B is a branch metric calculation portion, and 2B is provided in a side of the path metric calculating part of the outlet side of the 1B of branch metric calculation portion, and 4B is provided in a side of the relatively selection handling part of the outlet side of path metric calculating part 2B.
5 are provided in a side of the path metric memory of the input side of the outlet side of relatively selecting handling part 4B and path metric calculating part 2B, and the 6th, with the 1B of branch metric calculation portion, the remaining path memory of relatively selecting handling part 4B to be connected.
The 7th, the received signal input, received signal is from this end input.The 8th, the transmission line property input, the characteristic of transmission line for example the described tap coefficient of the foregoing description etc. from this end input.The 9th, export the output line in the remaining path of storage in the remaining path memory 6.
Secondly, corresponding to the contents processing among Figure 20, the working condition of this sequence estimation device is described.The 1B of branch metric calculation portion carries out BMG-I shown in Figure 20.
Then, path metric calculating part 2B input is carried out ADD-I shown in Figure 20 by the branch metric that the 1B of branch metric calculation portion calculates.
Relatively select handling part 4B input by a plurality of path metrics that path metric calculating part 2B calculates, carry out CMP-I, SEL.
In addition, in handling, the remaining path that is obtained is deposited in the remaining path memory 6 by the SEL that relatively selects handling part 4B to carry out.In addition, the path metric in remaining path is also deposited in the path metric memory 5.
More than Shuo Ming DFSE only notes the state of two memories in memory L, so indeterminate the best.Therefore, MLSE is produced deterioration in characteristics.For example, under the non-minimum phase condition in Figure 14 (b), the signal component of the memory of noting in DFSE reduces significantly than the signal component suitable with other memory, therefore produces the selection mistake in remaining path.And, in case produce the selection mistake in remaining path, just produce the phenomenon that this mistake is carried out " mistake transmission " continuously, there is the significantly problem of deterioration in characteristics that produces.
On the other hand, though MLSE has desirable part on characteristic, operand is too many, so there is unpractical problem.
Disclosure of an invention
The objective of the invention is to the estimation that reduces operand, carry out more accurate transmission sequence.
Serial estimating method of the present invention is a kind of viterbi algorithm that utilizes, during reaching second mulitpath that how to shift before constantly, select respectively and above-mentioned second constantly the corresponding remaining path of a plurality of data combination states from representing first constantly the data combination state, characteristic according to received signal and transmission line, the serial estimating method of the transmission burst that estimation sends from transmitter side, this serial estimating method may further comprise the steps: calculate from first constantly the state and reach second constantly the first path metric calculation step of path metric in path of state; Deriving change reaches second constantly the alternative routing of alternative routing of a part in path in path of state from above-mentioned first constantly the state and derives step; Calculating is derived the second path metric calculation procedure of the path metric of the alternative routing of deriving in the step at above-mentioned alternative routing; According to result of calculation in the above-mentioned first path metric calculation step and the result of calculation in the above-mentioned second path metric calculation procedure, revise from above-mentioned first constantly the state and reach second constantly the path modification step in path in path of state; Path by in above-mentioned path modification step, having revised, from reach second constantly a plurality of paths of state from above-mentioned first constantly the state, determine to reach above-mentioned second constantly the remaining path determining step in remaining path of state; After carrying out the above-mentioned first path metric calculation step, above-mentioned alternative routing derivation step, the above-mentioned second path metric calculation procedure, above-mentioned path modification step and above-mentioned remaining path determining step, to each state in a plurality of data combination states in above-mentioned second moment, from the remaining path that a plurality of data combinations in above-mentioned second moment are determined respectively, select the final path of final path to select step according to the path metric in this remaining path; And the burst that estimation obtains from the final path of selecting at above-mentioned final path to select the step is as the estimation steps that sends burst.
In addition, to derive step be to reach second constantly the step of a part in path in path of state according to the transmission line property change from above-mentioned first constantly the state for above-mentioned alternative routing.
In addition, it is to reach from above-mentioned first constantly the state in second constantly the branch in path of state constituting that above-mentioned alternative routing is derived step, change with respect to above-mentioned second constantly particular moment the branch after in the past step.
In addition, above-mentioned particular moment is the moment of determining according to degree of influence to received signal.
In addition, it is in the branch in the path that constitutes the state that reached for second moment from the state in above-mentioned first moment that above-mentioned alternative routing is derived step, changes the step of a plurality of branches.
In addition, above-mentioned remaining path determining step is according to reaching above-mentioned second constantly the path metric of mulitpath of state from above-mentioned first constantly the state, determines to reach above-mentioned second constantly the step in remaining path of state by the path of having revised in above-mentioned path modification step.
In addition, the difference calculation procedure that also has path metric, in this step, obtain path by in above-mentioned path modification step, having revised, reach difference between above-mentioned second constantly the path metric of mulitpath of state from above-mentioned first constantly the state, the above-mentioned second path metric calculation procedure be according to the difference of the path metric of in the difference calculation procedure of above-mentioned path metric, calculating in the past and in the above-mentioned first path metric calculation step, calculate reach above-mentioned second constantly the path metric in path of state from above-mentioned first constantly the state, calculate the step of the path metric of above-mentioned alternative routing.
In addition, definite result according to above-mentioned remaining path determining step, carry out the above-mentioned first path metric calculation step, above-mentioned alternative routing derivation step, the above-mentioned second path metric calculation procedure, above-mentioned path modification step and above-mentioned remaining path determining step once more, it is to select the step of final path from the remaining path that the result who carries out once more obtains that above-mentioned final path is selected step.
Sequence estimation device of the present invention is a kind of viterbi algorithm that utilizes, during reaching second mulitpath that how to shift before constantly, select respectively and above-mentioned second constantly the corresponding remaining path of a plurality of data combination states from representing first constantly the data combination state, characteristic according to received signal and transmission line, the sequence estimation device of the transmission burst that estimation sends from transmitter side, this sequence estimation device comprises with the lower part: calculate from above-mentioned first constantly the state reach second constantly state the path path metric and will reach second constantly the path metrics calculator of path metric of part alternative routing after changing in path in path of state from above-mentioned first constantly the state; According to the result of calculation of above-mentioned path metrics calculator, revise from above-mentioned first constantly the state and reach second constantly the path modification device in path in path of state; Path by in above-mentioned path modification device, having revised, from reach second constantly a plurality of paths of state from above-mentioned first constantly the state, determine to reach above-mentioned second constantly the remaining path determiner in remaining path of state; From the remaining path of above-mentioned second constantly a plurality of data combinations being determined respectively by above-mentioned remaining path determiner, select the final path selector of final path according to the path metric in this remaining path; And the burst that estimation obtains from the final path selected above-mentioned final path selector is as the estimation device that sends burst.
The simple declaration of accompanying drawing
Fig. 1 is the flow chart of the processing sequence of the serial estimating method among the expression embodiment 1.
The figure in the remaining path when Fig. 2 is expression moment k=3.
Fig. 3 common path that to be expression make each state of moment k=4 and the figure of alternative routing.
The figure in the remaining path when Fig. 4 is expression moment k=4.
Fig. 5 is that expression reaches the constantly common path of the state 00 of k=5 and the figure of alternative routing.
The figure in the remaining path when Fig. 6 is expression moment k=6.
Fig. 7 is the flow chart that expression sends the estimation process order of sequence.
Fig. 8 is the serial estimating method of expression among the embodiment 1 and the analog result of the performance of the serial estimating method of conventional example.
Fig. 9 is the block diagram of the sequence estimation device among the embodiment 2.
Figure 10 is the flow chart of the processing sequence of the serial estimating method among the expression embodiment 3.
Figure 11 is the analog result of the relation of expression number of repetition and bit error rate.
Figure 12 is the figure of the transmission line pattern of expression 6 taps.
Figure 13 is the figure of the transmission line pattern of expression 4 taps of adopting the serial estimating method among the embodiment 1.
Figure 14 is the electrical power distribution map of the tap coefficient of expression 4 taps
Figure 15 is the figure of difference that represents the contents processing of MLSE and DFSE briefly.
Figure 16 is the figure of the pattern of expression transmission line and sequence estimation.
Figure 17 is the figure of the best judgment model of expression.
Figure 18 is the figure of the transmission line pattern of expression 3 taps.
Figure 19 is lattice figure.
Figure 20 is the flow chart of the order of the existing serial estimating method of expression.
Figure 21 is the figure of expression with the final path of existing serial estimating method acquisition.
Figure 22 is the block diagram of existing sequence estimation device.
The optimal morphology of the usefulness that carries out an invention
Embodiments of the invention below are described.
Embodiment 1.
Flow process of the present invention when moment k has been shown among Fig. 1.In addition, for the difference of clear and definite and conventional example, surround the processing that increases newly with thick line.
The main difference point of the present invention and conventional example is as follows.
Except common branch/path metric, branch metric and the path metric during the remaining path of having calculated change in advance only in particular moment.
Path metric when not changing remaining path compares, and determines whether to revise remaining path.In addition, according to correction result, each branch is selected relatively to select the path metric of use in the processing.
The path metric of selecting under each state is compared selection to be handled.In addition, increase the processing of the difference of storage tolerance.
In order to illustrate in greater detail these differences, the transmission line pattern with among Figure 13 specifically describes working condition of the present invention.
Transmission line pattern among Figure 13 is the pattern that is applicable to the transmission line with tap coefficient shown in Figure 14 (b).That is, it is not have tap coefficient c 3～c 4The pattern that transmission line was suitable for.
Particularly in this form, the tap coefficient c of the non-minimum phase condition shown in Figure 14 (b) is described 5The transmission line pattern of value when bigger than other tap coefficient.Such transmission line modal representation for example through reflection wait and the received power of having passed through the received signal that receives after the stipulated time bigger than the received power of the received signal that directly receives from transmitter side.Imagination illustrates it later on the transmission line of above such transmission line modal representation.
Secondly, specify the processing sequence in each moment among this embodiment.
＜the processing of k=1～3 constantly 〉
At first, suppose that the data when k is negative constantly are known, states all when supposing k=0 all exist, and carry out following explanation.
From lattice Figure 19 as can be known, in order to reach s[1,00], s[0 is arranged, 00]/s[1,00] or s[0,01]/s[1,00] the path.Calculate their path metric, to comparing between the path metric that is obtained.Then, select the little path of path metric as remaining path.Here, select s[0,00]/s[1,00] the path as remaining path.
Equally, determine to reach s[1 singly, 10], s[1,01], s[1,11] remaining path before each state.
By carrying out same processing, also moment k=2, k=3 are determined to reach each state remaining path before.
Processing through such before moment k=3, obtains remaining path shown in Figure 2.In addition, poor (to call the poor of tolerance in the following text) of the path metric in the path metric in the remaining path of storage before the k=3 and selected path and the path of abandoning.
＜the processing of k=4 constantly 〉
Below, in the processing from k=1 to k=3, suppose state s[1,10] time remaining path be s[0,01]/s[1,10], show it be modified to s[0,00]/s[1,10] processing.
The calculating of the path metric in the path of [reaching state s[4.00]]
(the computing of branch metric: BMG-I)
Owing to become state s[4,00], thus can be state s[3,00] or state s[3,01].Therefore, as becoming state s[4,00] path of usefulness, the s[0 that the remaining path when existing with k=3 shown in Figure 2 interrelates, 01]/s[1,10]/s[2,01]/s[3,00]/s[4,00] path and s[0,01]/s[1,10]/s[2,11]/s[3,01]/s[4,00] such two paths in path.
At moment k=4, from this path s[0,01]/s[1,10]/and s[2,01]/s[3,00]/s[4,00] candidate of the transmission sequence of acquisition " 000101 ", at moment k=4, from this path s[0,01]/s[1,10]/s[2,11]/and s[3,01]/s[4,00] obtain the candidate of the transmission sequence of " 001101 ".
(calculating of change branch metric and the derivation of alternative routing: BMG-II)
Secondly, obtained to change in the candidate who sends sequence and tap coefficient c 5The transmission sequence of suitable value.
Send by change in the candidate of sequence and tap coefficient c 5Suitable value can obtain " 000100 " and " 001100 ".Should " 000100 " be the transmission sequence that has changed after the above-mentioned transmission sequence " 000101 ", " 001100 " be the transmission sequence that has changed after the above-mentioned transmission sequence " 001101 ".Here said change is meant that the specific value " 0 " that will be positioned at sends sequence changes to " 1 ", maybe will be positioned at the specific value " 1 " that sends sequence and change to " 0 ".
The path suitable with the transmission sequence " 000100 " that has changed is s[0,00]/and s[1,10]/s[2,01]/s[3,00]/[4,00], the path suitable with the transmission sequence " 001100 " that has changed is s[0,00]/s[1,10]/s[2,11]/s[3,01]/[4,00].The path that will obtain like this is called alternative routing later on.Concerning this alternative routing, before changing path is called common path.
With the s[4.00 that reaches that obtains like this] 4 paths be shown in Fig. 3 (a).In addition, in Fig. 3, wave molding is the branch that produces owing to change.As shown in Figure 3, we can say that alternative routing is with branch after changing the path of common path in past particular moment.
Secondly, utilize to send sequence " 000101 " and " 001101 ", after changing transmission sequence " 000100 " and " 001100 ", calculate branch metric.The computational methods of this branch metric can adopt the branch metric calculation method identical with existing method to calculate.
(the computing of path metric: ADD-I)
Then, calculate the common path and the path metric of alternative routing.
Branch metric during k=4 by will newly calculating is added in the path metric before the k=3 that calculates and storing over, can obtain the path metric in common path.
(the computing of the path metric of alternative routing: ADD-II)
In addition, calculate the path metric of alternative routing by following main points.
Alternative routing s[0,00]/s[1,10]/s[2,01]/s[3,00]/[4,00] and path s[0,01]/s[1,10]/s[2,01]/s[3,00]/[4,00] different part is the s[0 of branch, 00]/s[1,10] and the s[0 of branch, 01]/s[1,10].
Therefore, by will be, 00] from s[0/s[1,10] path metric deduct s[0,01]/s[1,10] difference behind the path metric (hereinafter referred to as path metric poor) is added to path s[0,01]/s[1,10]/s[2,01]/s[3,00] in the path metric/[4,00], can try to achieve the path metric of alternative routing.As the path s[0 of usefulness here, 00]/s[1,10] and path s[0,01]/s[1,10] the difference of path metric be the poor of path metric from the path of the having abandoned path metric that deducts remaining path, the value that this value stores after using and calculating when moment k=1.
According to path s[0,01]/s[1,10]/s[2,11]/s[3,01]/s[4,00] path metric, by to another alternative routing s[0,00]/s[1,10]/s[2,11]/s[3,01]/s[4,00] carry out same processing, calculate path metric.
Because the difference with the path metric of calculating in the past stores like this, be used for the calculating of alternative routing in the future, so can reduce the operand that the path metric of calculating alternative routing is used.
Through above processing, can calculate the common path and the path metric of alternative routing.In this embodiment, can calculate 4 path metrics of total.These 4 path metrics are temporarily stored.
The calculating of the path metric in [reaching state s[4.10] path before]
(calculating of the computing of branch metric: BMG-I, branch metric after changing and the derivation of alternative routing are handled: BMG-II)
The same with above-mentioned processing, reach state s[4.10] before Path Determination.
Owing to become state s[4,10], thus can be state s[3,00] or state s[3,01].Therefore, as becoming state s[4,10] path of usefulness, have s[0,01]/s[1,10]/s[2,01]/s[3,00]/s[4,00] path and s[0,01]/s[1,10]/s[2,11]/s[3,01]/s[4,00] such two paths in path.
At moment k=4, from this path s[0,01]/s[1,10]/s[2,01]/s[3,00]/and s[4,00] obtain the candidate of the transmission sequence of " 100101 ", at moment k=4, from path s[0,01]/s[1,10]/and s[2,11]/s[3,01]/s[4,10] obtain the candidate of the transmission sequence of " 101101 ".
Then, send in the candidate of sequence and tap coefficient c by change 5Suitable value obtains " 100100 " and " 101100 ".
To above transmission sequence " 100101 " and " 101101 ", after changing transmission sequence " 100100 " and " 101100 ", calculate branch metric.
Make again and this suitable alternative routing of transmission sequence after changing.Alternative routing is s[0,00]/s[1,10]/s[2,01]/s[3,00]/s[4,00] path and s[0,00]/s[1,10]/s[2,11]/s[3,01]/s[4,10] the path.
With the state that the reaches s[4 that obtains like this, 10] 4 paths be shown in Fig. 3 (b).
(computing of the path metric in common path and the path metric of alternative routing: ADD-I, ADD-II)
Then, calculate reach state s[4,10] common path and the path metric of alternative routing.The computational methods of path metric and state s[4,00] identical the time.
Through above such processing, can calculate the path metric of remaining path and alternative routing.Can calculate in this embodiment and amount to 4 path metrics.
[reaching state s[4.01] before the path and reach state s[4.11] calculating of the path metric in before path]
By with above identical processing, reach state s[4.01] calculating of the path metric in before path and reach state s[4.11] calculating of the path metric in before path.
By this processing, as reaching s[4.01] the path, obtain 4 paths shown in Fig. 3 (c), as reaching s[4.11] the path, obtain 4 paths shown in Fig. 3 (d).
(the correcting process in path: COR)
(a) selection of minimal path tolerance: COR-I
Secondly, select the path metric of numerical value minimum 4 path metrics of under each state, calculating.In this example, to state s[4.00] selection path s[0,00]/s[1,10]/s[2,01]/s[3,00]/and s[4,00], to state s[4.10] and selection path s[0,01]/s[1,10]/s[2,11]/and s[3,01]/s[4,10], to state s[4.01] and selection path s[0,00]/and s[1,10]/s[2,01]/s[3,10]/s[4,01], to state s[4.11] selection path s[0,00]/s[1,00]/s[2,10]/and s[3,11]/s[4,11], it is described later on.
(b) derivation of Dui Ying state: COR-II
Whether the path of these selections when reverse, is confirmed between the selected path by identical state.Specifically, carry out following processing.
In selected path, reach state s[4.00] the path, reach state s[4.10] the path, reach state s[4.01] the path all pass through state s[1.10], if but look to reach state s[1.10] process, then in these paths, have and have s[0,00]/s[1, the path of branch 10] and have s[0,01]/s[1,10] the path of branch.
(c) correction in remaining path: COR-III
Therefore, from this 3 paths, select the path of path metric minimum, as one man revise s[1,10 with selected path] before branch.The method of this correction is following carries out.
Suppose reaching state s[4.00] the path, reach state s[4.10] the path, reach state s[4.01] the path in, the path metric reckling is for reaching state s[4.01] the path, owing to reaching state s[4.01] the path in s[0 is arranged, 00]/s[1,10] so branch is with the s[1 in the dative composition 19,10] the remaining branch that interrelates is modified to s[0,00]/and s[1,10].
(d) with the selection of revising the path metric that accompanies in the path: COR-IV
At this constantly, the remaining branch during with moment k=1 is defined as s[0,00]/s[1,10].
Then, from the s[4.00 that reaches of temporary transient preservation] the path metric of 4 paths select s[0 is arranged, 00]/s[1,10] path s[0,00]/s[1,10]/s[2,01]/s[3,00]/s[4,00] path metric and path s[0,00]/s[1,10]/s[2,11]/s[3,01]/s[4,00] path metric.
(the comparison process of path metric: CMP-I)
Secondly to selected path s[0,00]/s[1,10]/s[2,01]/s[3,00]/s[4,00] path metric and path s[0,00]/s[1,10]/s[2,11]/s[3,01]/s[4,00] path metric compare.By getting poor (that measures is poor) between the selected path metric, carry out the comparison of path metric.
(the stores processor of the difference of path metric: CMP-II)
In the comparison process of path metric, poor (tolerance poor) of the path metric of calculating is stored, give over to later processing and use.
(select to handle: SEL)
Then, in path s[0,00]/s[1,10]/s[2,01]/s[3,00]/s[4,00] path metric and path s[0,00]/s[1,10]/s[2,11]/s[3,01]/s[4,00] path metric in, selecting path metric is that the path of reckling is as reaching s[4,00] remaining path.Then, select, upgrade path of storing so far according to this.
To each state s[4,10], s[4,01], s[4,11] also carry out the processing of above CMP-I, CMP-II and SEL.
Through above processing, as reaching state s[4,00] before remaining path, obtain s[0,00]/s[1,10]/s[2,01]/s[3,00]/s[4,00]; As reaching state s[4,10] before path, obtain s[0,00]/s[1,10]/s[2,11]/s[3,01]/s[4,10]; As reaching state s[4,01] before path, obtain s[0,00]/s[1,10]/s[2,01]/s[3,10]/s[4,01]; As reaching state s[4,11] before path, obtain s[0,00]/s[1,00]/s[2,10]/s[3,11]/s[4,11].
The remaining path that obtains during k=4 is shown in Fig. 4.
Processing during＜moment k=5 〉
(computing of the calculating of branch metric: BMG-I, BMG-II and path metric: ADD-I, ADD-II)
The same during also with k=4 about moment k=5, through the computing of branch metric and the computing of path metric, calculate the common path that reaches before each state and the path metric of alternative routing.
When moment k=5, as reaching state s[5,00] the path, have s[0,00]/s[1,10]/s[2,01]/s[3,00]/s[4,00]/s[5,00] path and s[0,00]/s[1,10]/s[2,01]/s[3,10]/s[4,01]/s[5,00] the path.
Then, obtain s[0,10 from these paths]/s[1,11]/s[2,01]/s[3,00]/s[4,00]/s[5,00] alternative routing and s[0,10]/s[1,11]/s[2,01]/s[3,10]/s[4,01]/s[5,00] alternative routing.In addition, the same when obtaining order that alternative routing uses with k=4, so explanation is omitted.
Reach state s[5,00] 4 paths as shown in Figure 5.
Thereby, calculate the path metric of this 4 paths.
Also obtain 4 paths respectively about other state, calculate the path metric of this 4 paths.
(correcting process: COR)
(a) selection of minimal path tolerance: COR-I
From 4 path metrics of under each state, calculating, select the path metric of numerical value minimum.
Here, in this embodiment, as reaching state s[5,00] before path, select s[0,00]/s[1,10]/s[2,01]/s[3,10]/and s[4,01]/s[5,00], as reaching state s[5,10] before path, select s[0,10]/and s[1,11]/s[2,01]/s[3,10]/and s[4,01]/s[5,10].
In addition, as reaching state s[5,01] before path, select s[0,11]/s[1,01]/s[2,10]/s[3,11]/and s[4,11]/s[5,01], as reaching state s[5,11] before path, select s[0,00]/and s[1,00]/s[2,10]/s[3,11]/s[4,11]/s[5,11].
(b) derivation of Dui Ying state: COR-II
According to the path of these selections, whether confirm between the selected path by identical state.Specifically, carry out following processing.
In selected path, reach state s[5.00] the path and reach state s[5.10] the path all pass through s[2.01], if but look to reach s[2,01] process, then one by s[0,00]/s[1,10]/s[2,01], another person is by s[0,10]/s[1,11]/s[2,01].
In addition, reach state " 01 " the path and reach state " 11 " the path all pass through s[2.10], if but look to reach s[2,10] process, then a paths is by s[0,11]/s[1,01]/s[2,10], another paths is by s[0,00]/s[1,00]/s[2,10].
(c) correction in remaining path: COR-III
Reaching state " 00 " the path and reach state " 10 " the path in, select the little path of path metric.Here, reaching state " 00 " the path metric ratio in path reach state " 10 " and the path metric in path little.
In the case, and reach state " 00 " the path as one man to reaching state " 10 " and the path also to revise so that by s[0,00]/s[1,10]/s[2,01].In addition, do not revise the state of reaching " 00 " the path.
In addition, and reach state " 11 " the path as one man revise the state of reaching " 01 " and the path so that by s[0,00]/s[1,00]/s[2,10].Do not revise the state of reaching " 11 " the path.In addition, reach state here " 11 " the path metric ratio in path reach state " 01 " and the path metric in path little.
(d) selection of the path metric that carries out according to correction result: COR-IV
Correcting process by remaining path reaches state " 10 " the path and reach state " 01 " the path revised, so select the path metric in these paths according to correction result
Then, from the s[5 that reaches of temporary transient preservation, 10] the path metric of 4 paths in, select to have s[0,00]/s[1,10]/s[2,01] path s[0,00]/and s[1,10]/s[2,01]/s[3,10]/s[4,01]/s[5,10] path metric and path s[0,00]/and s[1,10]/s[2,01]/s[3,00]/s[4,00]/s[5,10] path metric.
In addition, about reaching state " 01 " the path, from the s[5 that reaches of temporary transient preservation, 01] the path metric of 4 paths in, select to have s[0,00]/s[1,00]/s[2,10] path s[0,00]/s[1,00]/s[2,10]/s[3,01]/s[4,11]/s[5,01] path metric.
(comparison process: CMP-I and stores processor: CMP-II)
Secondly, to selected path s[0,00]/s[1,10]/s[2,01]/s[3,10]/s[4,01]/s[5,10] path metric and path s[0,00]/s[1,10]/s[2,01]/s[3,00]/s[4,00]/s[5,10] path metric compare.The difference of the path metric of calculating in the time of in addition, will carrying out this comparison stores.
About reaching state " 01 " the path owing to just s[0 is arranged, 00]/s[1,00]/s[2,10] path s[0,00]/s[1,00]/s[2,10]/s[3,01]/s[4,11]/s[5,01], so do not need comparison.
(select to handle: SEL)
Then, in path s[0,00]/s[1,10]/and s[2,01]/s[3,10]/s[4,01]/s[5,10] path metric and path s[0,00]/s[1,10]/and s[2,01]/s[3,00]/s[4,00]/s[5,10] in the path metric, select the little path of path metric as reaching s[5,10] remaining path.
To each state s[5,00], s[5,01], s[5,11] also carry out the processing of above CMP-I, CMP-II and SEL.Correctly say, for reaching the state that does not carry out path modification " 00 " the path or reach state " 11 " and the path, do not need the specific processing that compares and select, in the selection of minimal path tolerance is handled, will be defined as remaining path as the minimum path of selecting of path metric.
The remaining path of determining like this is shown in Fig. 6.
＜later the processing of k=6 constantly 〉
After moment k=6, identical processing when carrying out with k=5 determines to reach the remaining path before each state.
Carry out above processing constantly at each.With each processing vague generalization constantly be shown in Fig. 1.The processing in each moment is corresponding with the content of processing shown in Figure 1.
The estimation process of＜transmission sequence 〉
Till a frame partly finishes, carry out processing shown in Figure 1, compare between the path metric to the remaining path of acquisition under each state at the eleventh hour, select the path of path metric minimum.The path of this selection is exactly a final path.
Taking out the transmission sequence from this final path, is the transmission sequence that transmitter side sends with the transmission sequence estimation that is obtained.
The flow process of the estimation process of such transmission sequence has been shown among Fig. 7.
By carrying out above processing, can estimate the transmission sequence that transmitter side sends at receiver side.
Analog result when the processing of adopting this embodiment has been shown among Fig. 8.
△ among Fig. 8 represents by carrying out the processing of this embodiment, the bit error rate the when bit error rate when estimation sends sequence, are represented to utilize existing DFSE estimation to send sequence.Bit error rate when in addition, zero expression utilizes existing MLSE estimation to send sequence.
In Fig. 8, the longitudinal axis is represented bit error rate, and transverse axis is represented average E b/ N oRatio.Average E b/ N oThan the mean value of the signal energy of representing each and the ratio of Carrier To Noise Power Density.
In this simulation, condition is: identical with transmission line pattern shown in Figure 12, and the electrical power of each tap coefficient is identical, the decay transmission line that is rayleigh distributed.In addition, as 4 states, as 32 states, MLSE of conventional example handles it as 4 states with it by present embodiment with the status number among the memory L for the DFSE of conventional example.
From analog result shown in Figure 8 as can be known,, compare, can reduce bit error rate significantly with the DFSE of conventional example by carrying out the processing of present embodiment.
In addition, check that computing is taken temperature, suppose operand with the DFSE of conventional example as 1, the operand of the MLSE of conventional example is 8 so, and the operand of present embodiment is 2, and operand is reduced significantly than existing MLSE.
As from the foregoing, the processing of present embodiment can reduce operand, can send the estimation of sequence again more accurately.Why can send the estimation of sequence so accurately, be because, compare selection, and revise the path not only to remaining path but also alternative routing is carried out the calculating of path metric.By carrying out the correcting process in this path, constantly different in the past paths can be selected and correct path can be modified to, can carry out the high estimation of reliability.
In addition, though adopted tap coefficient c in this embodiment 5The big transmission line pattern of influence, but also can adopt other transmission line pattern.In addition, in this embodiment owing to adopted tap coefficient c 5The big transmission line pattern of influence, so changed in the transmission sequence and tap coefficient c 5Corresponding value is if but the influence of other tap coefficient when big is necessary the position of being changed corresponding to the big tap coefficient change of this influence.
In addition, in this manual, transmission line comprises any one of wired transmission line and wireless transmission link.
In addition, in this embodiment, be that example is illustrated with " following feedback (ト レ one ス バ Star the Network) " formula of taking out estimation transmission sequence after the processing of carrying out a frame.In addition,, be called the method for " memory replacing " in addition, adopt the method for " memory replacing ", also can estimate the transmission sequence as the estimation process that sends sequence.In addition, can also adopt other method to carry out the processing of sequence estimation.This point, later example too.
Estimation steps among the present invention comprises the method for these " following feedback ", the method for " memory replacing " etc.
Embodiment 2.
This embodiment shows the sequence estimation device of the processing among the embodiment that carries out the front.
Fig. 9 is the block diagram that carries out the sequence estimation device of the processing among the embodiment of front.
Among the figure, 1A is a branch metric calculation portion, and 2A is provided in a side of the path metric calculating part of the outlet side of the 1A of branch metric calculation portion, and 3 are provided in a side of the remaining path modification portion of the outlet side of path metric calculating part 2A.4A is provided in a side of the relatively selection handling part of the outlet side of remaining path modification portion 3,5 are provided in a side of the memory of the input side of the outlet side of relatively selecting handling part 4A and path metric calculating part 2A, and this memory 5 has the function as the memory of the difference of the memory of store path tolerance and storage tolerance.Among the figure, memory 5 is recited as the difference memory of path metric storage metrics.
The 6th, with the 1A of branch metric calculation portion, path metric calculating part 2A, remaining path modification portion 3 and the remaining path memory of relatively selecting handling part 4A to be connected.
The 7th, the received signal input, received signal is from this end input.The 8th, the transmission line property input, the characteristic of transmission line for example the described tap coefficient of the foregoing description etc. from this end input.The 9th, export the output line in the remaining path of storage in the remaining path memory 6.
10 are provided in a side of the final path determination portion of the outlet side of the difference memory 5 of path metric storage metrics and remaining path memory 6.11 estimations that are provided in a side of the outlet side of final path determination portion 10 send sequence taking-up portion.
Secondly, the working condition of this sequence estimation device is described accordingly with the contents processing among Fig. 1.The 1A of branch metric calculation portion carries out BMG-I shown in Figure 1, BMG-II.At this moment, according to the characteristic of the transmission line of importing from transmission line property input 8, determine that change sends the value of which part of sequence.For example, as described in embodiment 1, at tap coefficient c 5The situation of the big transmission line of influence under, change sends in the sequence and tap coefficient c 5Corresponding value.
Then, path metric calculating part 2A input is carried out ADD-I shown in Figure 1, ADD-II by the branch metric that the 1A of branch metric calculation portion calculates.In addition, when path metric calculating part 2A calculates the path metric of alternative routing in ADD-I, ADD-II, read in the difference and the remaining path that is stored in the remaining path memory 6 of the tolerance in the difference memory 5 that is stored in the path metric storage metrics, calculate the path metric of alternative routing.
The path metric in the common path that 3 inputs of remaining path modification portion are calculated by path metric calculating part 2A and the path metric of alternative routing carry out COR-I shown in Figure 1, COR-II, COR-III, COR-IV.Remaining path modification portion 3 deposits revised path in the remaining path memory 6 in after having revised the path.
A plurality of path metrics of relatively selecting handling part 4A input to be selected by remaining path modification portion 3 carry out COM-I, COM-II, SEL.In the processing of COM-II, the difference of relatively selecting handling part 4A to measure deposits in the difference memory 5 of path metric storage metrics.In addition, in processing, the remaining path that is obtained is deposited in the remaining path memory 6 by the COM-II that relatively selects handling part 4A to carry out.In addition, the path metric in remaining path also deposits in the difference memory 5 of path metric storage metrics.
Above handling part each before a frame end carries out processing shown in Figure 1 constantly.
If the processing of frame part finishes, then final path determination portion 10 is from the path metric in the remaining path that difference memory 5 input of path metric storage metrics obtains corresponding to each state of last moment, and from remaining path that remaining path memory 6 inputs obtain corresponding to each state.
Final path determination portion 10 is by selecting the path of path metric minimum as final path from the remaining path of being imported.
Estimation sends sequence taking-up portion 11 and take out the transmission sequence from the final path of being determined by final path determination portion 10, is the transmission sequence that transmitter side sends with the transmission sequence estimation that is obtained.
Sequence estimation device among this embodiment can reduce operand by carrying out above processing, can carry out the estimation of more accurate transmission sequence again.Why can send the estimation of sequence so accurately, be because not only to remaining path but also alternative routing is carried out the calculating of path metric, and according to the result of calculation correction path of path metric.By carrying out the correcting process in this path, constantly different in the past paths can be selected and correct path can be modified to, can carry out the high estimation of reliability.
In addition, in this embodiment,, show final path determination portion 10 and estimation and send sequence taking-up portion 11 as the circuit structure of the estimation process that sends sequence.To utilize this structure to be called " following feedback " formula in the method for having carried out frame processing back taking-up estimation transmission sequence.In addition,, be called the method for " memory replacing " in addition, adopt the method for being somebody's turn to do " memory replacing ", also can estimate the transmission sequence as the estimation process that sends sequence.At this moment the part of circuit structure is different with present embodiment.
In addition, the estimation steps among the present invention comprises the method for carrying out above-mentioned " following feedback " and carries out in the method for " memory replacing " any.
Embodiment 3.
Among the embodiment in front, be conceived to a tap, made alternative routing., exist under a plurality of situations, also can be conceived to a plurality of taps, make alternative routing in the big tap of influence.
In the case, because alternative routing is done manyly,, better estimate but can carry out precision so treating capacity increases.
Embodiment 4.
Among the embodiment in front, show each situation of constantly only carrying out once processing shown in Figure 1 before a frame partly finishes.
, also can utilize the remaining path that obtains by primary processing to carry out processing shown in Figure 1 once more.
In correcting process shown in Figure 1,, owing to revise the path of each state when reaching k=1,, do not carry out the calculating of the branch metric that carries out according to revised path or the calculating of path metric etc. in the moment of k=1 for example in the moment of k=4.
Therefore,, can handle, better estimate so can carry out precision according to further revised path by carrying out the processing of the BMG-I～SEL among Fig. 1 once more.
That is, handle according to order shown in Figure 10.In addition, when carrying out reprocessing, do not carry out CMP-I.
The number of repetition how many times can, but as shown in figure 11, if carry out to a certain degree repetition, then bit error rate can be restrained.Therefore, can determine only number of repetition according to this characteristic.
Analog result when the processing of carrying out this embodiment has been shown among Fig. 8.◇ among Fig. 8 is the analog result when using the remaining path of being determined by primary processing to repeat processing shown in Figure 1 as shown in figure 10.Number of repetition be shown be 3 times situation in this analog result.
Like this, by carrying out reprocessing, can improve error rate as can be known, better estimate thereby can carry out precision.
Embodiment 5.
Among the embodiment in front, having illustrated does not have tap coefficient c 3, c 4The transmission line pattern, but also can adopt tap coefficient c 3, c 4The transmission line pattern.
But, except tap coefficient c 0～c 2Also exist in addition under the situation of a plurality of tap coefficients, as described below, revise the path if make, then can reduce treating capacity.
For change with except tap coefficient c 0～c 2The value of the signal of storing in the memory of the tap coefficient correspondence that has the greatest impact in a plurality of tap coefficients in addition makes and revises the path.
For example, tap coefficient c in tap coefficient 3Have the greatest impact the time, the change with tap coefficient c 3The value of the signal of storing in the corresponding memory makes and revises the path.
In addition, the tap coefficient that what is called has the greatest impact is meant the tap coefficient of coefficient value for maximum.
Embodiment 6.
In above sequence estimation device, tap coefficient is fixed, but also can be according to the transmission line property from 8 inputs of transmission line property input, the number of change tap coefficients value and tap coefficient.
Owing to have the tap coefficient change step or the device of change tap coefficients value and tap coefficient, no matter which type of transmission line can both send the estimation of signal to.In addition, even also can be corresponding after transmission line property changes.
The present invention is owing to as above consisting of, so have the effect of the following stated.
Because serial estimating method of the present invention is a kind of viterbi algorithm that utilizes, from representing first Data assembled state is constantly selected in reaching the mulitpath that how to shift before second moment The remaining path corresponding with above-mentioned second constantly a plurality of data assembled state respectively is according to reception The characteristic of signal and transmission line, the sequence of the transmitted signal sequence that estimation sends from the sending side is estimated The calculation method, this serial estimating method may further comprise the steps: calculating reaches from the state in first moment The first path metric calculation step of the path metric in the path of the state in second moment; Derive and become More reach second constantly the change of a part in path of state from above-mentioned first constantly the state The alternative routing in path is derived step; Calculating is derived the change of deriving in the step at above-mentioned alternative routing The second path metric calculation procedure of the path metric in path more; According to above-mentioned first path metric Result of calculation in result of calculation in the calculation procedure and the above-mentioned second path metric calculation procedure, Correction reached for second constantly path modification step in path of state from above-mentioned first constantly the state Suddenly; Path by in above-mentioned path modification step, having revised, from from above-mentioned first constantly State reaches determines to reach above-mentioned second constantly the state in second constantly a plurality of paths of state The remaining path determining step in remaining path; Carrying out the above-mentioned first path metric calculation step Suddenly, above-mentioned alternative routing is derived step, the above-mentioned second path metric calculation procedure, above-mentioned path Revise after step and the above-mentioned remaining path determining step, to above-mentioned second constantly a plurality of Each state in the data assembled state is from a plurality of data combination difference to above-mentioned second moment In the remaining path of determining, select the last of final path according to the path metric in this remaining path The Path selection step; And the way of escape of estimation from selecting at above-mentioned final path to select the step The burst that the footpath obtains is as the estimation steps of transmitted signal sequence, so can reduce computing Amount can be carried out again the estimation of more accurate transmission sequence.
In addition, owing to above-mentioned alternative routing derivation step changes from above-mentioned according to transmission line property First constantly the state reaches second constantly the part in path of state, so can change root Lines feature change's path reportedly, it is better to carry out according to the transmission line of reality precision Send the estimation of sequence.
In addition, because above-mentioned alternative routing is derived step at the state of formation from above-mentioned first moment Reach in second constantly the branch in path of state, change is with respect to above-mentioned second constantly the spy Branch after regularly carving is so become by the branch that is conceived to after particular moment goes over More, can carry out the estimation that precision better sends sequence.
In addition, because the feature of above-mentioned particular moment is to determine according to degree of impact to received signal The moment, so by changing in the lump with degree of impact to received signal, can change docking The high branch of degree of impact that collects mail number can carry out that effect is higher, precision better sends sequence Estimation.
In addition, because above-mentioned alternative routing is derived step at the state of formation from above-mentioned first moment Reach in second constantly the branch in path of state and change a plurality of branches, so can revise a plurality of Branch can carry out the estimation that precision better sends sequence.
In addition, owing to above-mentioned remaining path determining step reaches according to the state from above-mentioned first moment To above-mentioned second constantly the path metric of mulitpath of state, determine by at above-mentioned path Revise the path of having revised in the step and reach above-mentioned second constantly the remaining path of state, so Path metric as metewand, can be sent the estimation of sequence more rightly.
In addition, because the difference calculation procedure of path metric is arranged, in this step, obtain by upper State the path revised in the path modification step, reach above-mentioned the from above-mentioned first constantly the state Difference between the path metric of the mulitpath of the state in two moment, above-mentioned second path metric Calculation procedure according to the path metric of in the difference calculation procedure of above-mentioned path metric, calculating in the past it Difference and the state from above-mentioned first moment of calculating in the above-mentioned first path metric calculation step reach To above-mentioned second constantly the path metric in path of state, the path of calculating above-mentioned alternative routing Tolerance is so can easily calculate the path metric of alternative routing.
In addition, owing to the definite result according to above-mentioned remaining path determining step, again carry out State the first path metric calculation step, above-mentioned alternative routing derivation step, the above-mentioned second path degree Amount calculation procedure, above-mentioned path modification step and above-mentioned remaining path determining step, above-mentioned Final path selects step to select the way of escape from the remaining path that the result who again carries out obtains Directly, so by carrying out reprocessing, can improve error rate, can carry out precision and better send The estimation of sequence.
Because sequence estimation device of the present invention is a kind of viterbi algorithm that utilizes, from representing first Data assembled state is constantly selected in reaching the mulitpath that how to shift before second moment The remaining path corresponding with above-mentioned second constantly a plurality of data assembled state respectively is according to reception The characteristic of signal and transmission line, the sequence of the transmitted signal sequence that estimation sends from the sending side is estimated Calculate device, this sequence estimation device comprises following part: calculate from the state in above-mentioned first moment Reach second constantly state the path path metric and will reach from above-mentioned first constantly the state To second constantly the road of path metric of part alternative routing after changing in path of state The footpath metric calculator; According to the result of calculation of above-mentioned path metrics calculator, revise from above-mentioned the One constantly state reaches second constantly the path modification device in path of state; By above-mentioned The path of having revised in the path modification device, from reaching above-mentioned second from above-mentioned first constantly the state Determine to reach above-mentioned second constantly the remaining path of state in a plurality of paths of state constantly Remaining path determiner; From by above-mentioned remaining path determiner to above-mentioned second constantly a plurality of numbers According to making up in the remaining path of determining respectively, select last according to the path metric in this remaining path The final path selector in path; And estimation is from selecting above-mentioned final path selector The burst that final path obtains is as the estimation device of transmitted signal sequence, so can reduce fortune The calculation amount can send the estimation of sequence more accurately.
The possibility of industrial utilization
As mentioned above, be applicable to,, send burst in the receiver side estimation according to the characteristic of received signal and transmission line being in the digital data transfer of representative with automobile telephone etc.Operand can be reduced, the estimation of more accurate transmission sequence can be carried out again.

Claims (7)

1. serial estimating method, it is to utilize viterbi algorithm, during reaching second mulitpath that how to shift before constantly, select respectively and above-mentioned second constantly the corresponding remaining path of a plurality of data combination states from representing first constantly the data combination state, characteristic according to received signal and transmission line, estimation is from the serial estimating method of the transmission burst of transmitter side transmission, and this serial estimating method is characterised in that and may further comprise the steps:
Calculating reaches second constantly the first path metric calculation step of path metric in path of state from first constantly the state;
Deriving change reaches second constantly the alternative routing of alternative routing of a part in path of state from above-mentioned first constantly the state and derives step;
Calculating is derived the second path metric calculation procedure of the path metric of the alternative routing of deriving in the step at above-mentioned alternative routing;
According to result of calculation in the above-mentioned first path metric calculation step and the result of calculation in the above-mentioned second path metric calculation procedure, revise from above-mentioned first constantly the state and reach second constantly the path modification step in path of state;
Path by in above-mentioned path modification step, having revised, from reach second constantly a plurality of paths of state from above-mentioned first constantly the state, determine to reach above-mentioned second constantly the remaining path determining step in remaining path of state;
After carrying out the above-mentioned first path metric calculation step, above-mentioned alternative routing derivation step, the above-mentioned second path metric calculation procedure, above-mentioned path modification step and above-mentioned remaining path determining step, to each state in a plurality of data combination states in above-mentioned second moment, from the remaining path that a plurality of data combinations in above-mentioned second moment are determined respectively, select the final path of final path to select step according to the path metric in this remaining path; And
The burst that estimation obtains from the final path of selecting at above-mentioned final path to select the step is as the estimation steps that sends burst.
2. serial estimating method according to claim 1, it is characterized in that: above-mentioned alternative routing is derived step and is reached from above-mentioned first constantly the state in second constantly the branch in path of state constituting, and change is with respect to above-mentioned second constantly the branch of particular moment after in the past.
3. serial estimating method according to claim 2 is characterized in that: above-mentioned particular moment is determined according to the timing of the big reflected wave of the signal electrical power of associated transmission lines pattern.
4. serial estimating method according to claim 1 is characterized in that: above-mentioned alternative routing is derived step in the branch in the path that constitutes the state that reached for second moment from the state in above-mentioned first moment, changes a plurality of branches.
5. serial estimating method according to claim 1, it is characterized in that: the difference calculation procedure that path metric is arranged, in this step, obtain path by in above-mentioned path modification step, having revised, reach difference between above-mentioned second constantly the path metric of mulitpath of state from above-mentioned first constantly the state
The above-mentioned second path metric calculation procedure according to the difference of the path metric of in the difference calculation procedure of above-mentioned path metric, calculating in the past and in the above-mentioned first path metric calculation step, calculate reach above-mentioned second constantly the path metric in path of state from above-mentioned first constantly the state, calculate the path metric of above-mentioned alternative routing.
6. according to each described serial estimating method of claim 1 to 5, it is characterized in that: according to definite result of above-mentioned remaining path determining step, carry out the above-mentioned first path metric calculation step, above-mentioned alternative routing derivation step, the above-mentioned second path metric calculation procedure, above-mentioned path modification step and above-mentioned remaining path determining step once more
Above-mentioned final path selects step to select final path from the remaining path that the result who carries out once more obtains.
7. sequence estimation device, it is to utilize viterbi algorithm, during reaching second mulitpath that how to shift before constantly, select respectively and above-mentioned second constantly the corresponding remaining path of a plurality of data combination states from representing first constantly the data combination state, characteristic according to received signal and transmission line, estimation is from the sequence estimation device of the transmission burst of transmitter side transmission, and this sequence estimation device is characterised in that and comprises with the lower part:
Calculating from above-mentioned first constantly the state reach second constantly state the path path metric and will reach second constantly the path metrics calculator of path metric of part alternative routing after changing in path in path of state from above-mentioned first constantly the state;
According to the result of calculation of above-mentioned path metrics calculator, revise from above-mentioned first constantly the state and reach second constantly the path modification device in path in path of state;
Path by in above-mentioned path modification device, having revised, from reach second constantly a plurality of paths of state from above-mentioned first constantly the state, determine to reach above-mentioned second constantly the remaining path determiner in remaining path of state;
From the remaining path of above-mentioned second constantly a plurality of data combinations being determined respectively by above-mentioned remaining path determiner, select the final path selector of final path according to the path metric in this remaining path; And the burst that estimation obtains from the final path selected above-mentioned final path selector is as the estimation device that sends burst.
CN97194349A 1997-03-04 1997-03-04 sequence estimation method Expired - Fee Related CN1100395C (en)

Priority Applications (5)

Application Number Priority Date Filing Date Title
CN97194349A CN1100395C (en) 1997-03-04 1997-03-04 sequence estimation method
AU73213/98A AU705414B2 (en) 1997-03-04 1997-03-04 Method of sequence estimation
PCT/JP1997/000652 WO1998039848A1 (en) 1997-03-04 1997-03-04 Serial estimating method
CA002253395A CA2253395C (en) 1997-03-04 1997-03-04 Method of sequence estimation
EP97903655A EP0902545A4 (en) 1997-03-04 1997-03-04 Serial estimating method

Applications Claiming Priority (3)

Application Number Priority Date Filing Date Title
CN97194349A CN1100395C (en) 1997-03-04 1997-03-04 sequence estimation method
PCT/JP1997/000652 WO1998039848A1 (en) 1997-03-04 1997-03-04 Serial estimating method
CA002253395A CA2253395C (en) 1997-03-04 1997-03-04 Method of sequence estimation

Publications (2)

Publication Number Publication Date
CN1217837A CN1217837A (en) 1999-05-26
CN1100395C true CN1100395C (en) 2003-01-29

Family

ID=25680629

Family Applications (1)

Application Number Title Priority Date Filing Date
CN97194349A Expired - Fee Related CN1100395C (en) 1997-03-04 1997-03-04 sequence estimation method

Country Status (2)

Country Link
CN (1) CN1100395C (en)
CA (1) CA2253395C (en)

Also Published As

Publication number Publication date
CA2253395A1 (en) 1998-09-11
CN1217837A (en) 1999-05-26
CA2253395C (en) 2002-10-22

Similar Documents

Publication Publication Date Title
CN1253790C (en) Display device and driving method thereof
CN1488209A (en) Multi-pass interference removal apparatus and multi-pass interference removal method
CN1320769C (en) Coder, decoder and data transfer systems
CN1108665C (en) Digital transmission system and method comprising product code combined with multidimensional modulation
CN1139191C (en) Turbo-code error correcting decoder, turbo-code error correcting decoding method, turbo-code decoder and system thereof
CN1201494C (en) Method and apparatus for decoding maximum posterior probaility
CN1296686A (en) Transmitter/receiver
CN1078779C (en) Diversity receiver
CN1433167A (en) OFDM communication equipment, OFDM communication method and OFDM communication program
CN1863184A (en) Signal decomposition apparatus and signal decomposition method
CN1224177C (en) Method for receiving multipath signals in radio communications system with CDMA and device for carrying out said method
CN1873625A (en) Method for automatic generating random excitation based on percentage of function coverage
CN1770864A (en) 4x4 discrete cosine transform rapid parallel device based on AVS and its method
CN1898723A (en) Signal decoding apparatus and signal decoding method
CN1890624A (en) Electronic device, control method thereof, host device, and control method thereof
CN1393066A (en) Receiving apparatus and receiving method for radio communication
CN1022591C (en) Address processor for signal processor
CN1925597A (en) Image processing apparatus, image processing method, and program
CN1647524A (en) Image conversion device and image conversion method
CN1605059A (en) Pipeline core in montgomery multiplier
CN1836372A (en) System estimation method, program, recording medium, system estimation device
CN1100395C (en) sequence estimation method
CN1218495C (en) Radio signal receiving system
CN1545683A (en) Data converting device and method, learning device and method, and program and recording medium
CN1708908A (en) Digital signal processing method, processor thereof, program thereof, and recording medium containing the program

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20030129

Termination date: 20150304

EXPY Termination of patent right or utility model