JP3171801B2 - Decoding method and apparatus using trace deletion method for Viterbi algorithm - Google Patents

Description

Description: BACKGROUND OF THE INVENTION [0001] 1. Field of the Invention [0002] The present invention relates to Viterbi.
2.) An algorithm-based decoding system, in particular, by removing the branch path from the previous state that enters the state without a connection to the next state, to optimize the decoding for the data. Survival pathway (surv
The present invention relates to a decoding method and apparatus for determining ivor path). 2. Description of the Related Art Generally, in digital communication, a signal transmitted from a transmission unit is changed to a signal different from an original signal due to the influence of noise imposed on a channel and received by a reception unit. Therefore, digital communication systems use error correction codes to reduce errors generated in the signal. [0003] The convolution of the error correction coding technique (convol
ution) TCM (Trel) obtained by coding and modulation techniques
The lis Coded Modulation) signal is decoded by a demodulator and a decoder implementing the Viterbi algorithm. The Viterbi algorithm not only detects TCM signals but also MLSE
(Maximum likelihood sequence estimation) is efficiently implemented, and is widely used for detecting convolutional codes. This Viterbi algorithm is an algorithm that basically performs maximum likelihood decoding and greatly reduces the required calculation amount using a trellis diagram. In this algorithm, only one surviving path exists in one state through a similarity comparison between a path input to each state and an input signal. This process is repeated at each stage of the trellis diagram. A Viterbi decoding device using an existing trace tracking method will be described with reference to FIGS. In FIG. 1, BMG (Branch Metric Geometry)
The nerator) unit 11 calculates a branch evaluation amount of the input data by a state transition from a state at a specific time to a state at the next time on a trellis diagram as in the example of FIG. , Tn shown on the horizontal axis in FIG. 2 indicate time points, respectively, and are "00", "10", "0" shown on the vertical axis.
“1” and “11” indicate states, respectively, so that each state corresponding to a specific point in time has a branch that can transition to the state at the next point in time, and the number of branches is determined by a trellis diagram. 11 calculates a branch metric corresponding to a branch that can transition from one state at a specific point in time, and an ACS (Add-Compare-Select) unit 12 that receives a branch metric applied from the BMG unit 11 generates Is determined, and a decision vector indicating a branch route added to the surviving route is output to the surviving memory unit 13. More specifically, the ACS unit 12 enters each state at the current point in time. The branch metric of the coming branch and the state metric corresponding to the surviving path at the time before being connected to this branch are added, and the state metric and the branch are added. By comparing the partial path evaluation amounts obtained by summing the partial path evaluation amounts with the valuation amounts, a partial path corresponding to the optimum partial path evaluation amount is selected as a surviving path for the corresponding state at the current time. This is the sum of the branch evaluation amounts of the branch routes that constitute the surviving route, where the “branch” is a route that can transition from the state at one time to the state at the next time, and the “branch route” is on the surviving route. In the case of a binary symmetric channel, for example, the surviving path is the path having the smallest partial path metric. Then, a surviving path corresponding to all states at the current time point is selected.Since the evaluation amount described above is known in relation to the Viterbi algorithm, a specific description thereof will be given. When a surviving path individually corresponding to the current state is determined, the ACS unit 12 outputs a determination vector indicating a branch path newly added to the surviving path to the surviving memory unit 13. When the decision vector of the decoding depth is input, the survival memory unit 13 storing the decision vector from the ACS unit 12 reversely traces the surviving path finally left based on the decision vector, and performs the reverse tracing process. The transmitted data is finally determined, and information corresponding to the determined data is output as decoded information, but the conventional decoder device as described above uses a decoding depth determination vector. If an error occurs when the final surviving route is determined by using the method of backtracking the surviving route finally left The final surviving path erroneously determined, therefore there is a problem of determining erroneously transmitted data. This is because the reverse tracking method for determining transmitted data is blocked at the decoding depth, and therefore, there is no additional constraint other than the decoding depth constraint when implementing an ideal Viterbi algorithm. Acts as a loss. [0007] In order to solve the above-mentioned problems, an object of the present invention is to eliminate the above-described blocking by removing branch paths that enter a state that is not connected to the next state. To provide a decoding method for preventing performance degradation due to. It is another object of the present invention to provide a decoding apparatus embodying the above-described method for removing a branch path entering a state not connected to the next state. [0008] In order to achieve the above-mentioned object of the present invention, a method of decoding input data using the Viterbi algorithm according to the present invention comprises the steps of (a) input data and grid-like decoding; Iteratively determining a surviving path corresponding to each state at each time point based on the figure, and (b) having no branch path connected to the next state among the surviving paths determined in step (a). The surviving route is searched for and the surviving route determined in step (a) is determined.
Branch route to the (n + 1) th state based on the road
An n-th state detection operation having no path is performed, and (n
-1) n-th detected from each of the states at the time
Route number information indicating the number of branch routes that advance to the state at the time
And generating a deletion branch path in the surviving path that Sagashidasa in step (c) (b), originating in step (b)
The method comprising the branch path to be deleted based on generated route number information is repeated while dates back to a previous state to not overlap with other surviving path having a branch path which is connected to the next state, (d) the step (c) by way comprising the steps of generating a decrypted data Ozuru the input data on the basis of the finally remaining survivor path
Become In order to achieve another object of the present invention,
The apparatus for decoding input data according to the present invention using the Viterbi algorithm repeatedly determines branch paths corresponding to each state at each time point based on the input data and the grid diagram, First means for generating a decision vector and the number of paths information, storing the decision vector generated by the first means, and outputting path information indicating each surviving path determined by the stored decision vector. Means for storing the number-of-paths information output from the first means and the path information output from the second means; and storing the first information from the stored path information.
The operation of deleting the information of the branch path corresponding to the path number information of the means is repeatedly performed while returning to the previous state while the branch path to be deleted does not overlap with another surviving path having the branch path connected to the next state, Third means for generating decoded data corresponding to the input data based on the path information of the surviving path finally left. Preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. The trace deletion method proposed by the present invention comprises: (1) determining a surviving path for each state at each point in time using the Viterbi algorithm; and (2) connecting to the next state among the determined surviving paths. Searching for a surviving path that does not have the determined branch path, and (3) deleting a branch path in the searched surviving path by using another branch path having a branch path connected to the next state. Repeating the previous state while not overlapping with the surviving path. Referring to FIG. 3 which embodies the trace deletion method, the BMG unit 11 performs the operation described with reference to FIG. 1 on the input data at each time on the grid diagram. The obtained branch evaluation amount is output.
The ACS unit 32 stores a state evaluation amount related to a surviving path individually responding to each state at the previous time, and includes n +
After adding the branch metric related to the branch from the n-th time that can transition to each state at the first time to the corresponding state metric, and comparing the path metric obtained from the result, the n + 1-th time A selection signal for selecting a branch path having an optimum path evaluation amount for each state is generated. The ASC unit 32 generates a decision vector indicating the selected branch path for each state, and outputs the generated decision vector to the surviving memory unit 33. ASC section 32
The number of paths indicating the number of branch paths connected to the state at the (n + 1) th time point in each state at the nth time point using the selection signal used for selecting the branch path newly added to the surviving path at the (n + 1) th time point Generate information. The generated path number information is supplied to the path number memory unit 34, and the path number memory unit 34 stores the supplied path number information. The survivor memory unit 33 stores the decision vectors supplied from the ACS unit 32, generates path information indicating a surviving path based on the stored decision vectors, and supplies the generated path information to the path number memory unit 34. . The number-of-paths memory unit 34 performs the trace deletion method presented by the present invention using the stored number-of-paths information and the path information supplied from the surviving memory unit 33. The operation of the path number memory unit 34 will be described with reference to FIG. 2 and its flowchart, FIG. Number of paths memory unit 3
4 searches for the state at the nth time point that is not connected to the state at the n + 1th time point by the surviving path based on the path number information (step 401). The number-of-paths memory unit 34 generates a path deletion signal for deleting information indicating the searched state at the n-th time point or the branch path from the (n-1) -th time point connected to the state (step 402). (Step 403). In step 404, based on the path number information corresponding to the state at the (n-1) th time point when the information of the branch path is deleted, the path number memory unit 34 has a branch path connected from that state to another state at the nth time point. Judge. If not, the path number memory unit 34 generates a path deletion signal for deleting information indicating the state at the (n-1) th time point or the branch path from the (n-2) th time point linked to the state (step 402). The information of the corresponding branch route is deleted from the route information according to the route deletion signal (step 40).
3). On the other hand, if it exists in step 404, the path number memory unit 34 stops the path deletion operation. For easy understanding, a case where the surviving path of each state finally determined at each time point is as shown in FIG. 2 will be described. The path number memory unit 34 searches for “00” in a state where there is no branch path connected to the next time point t5 in step 401 at time point t4, and performs steps 402 and 403 to return to “01” which was the state at the previous time point t3. The value of the corresponding path number information is reduced by one, and the path information corresponding to the branch path connected to the state “00” from the state “01” at the time t3 is deleted. After that, since there is a branch path connected from the state “01” at the time point t3 to another state “10” at the time point t4, the path number memory unit 34 stops the path deletion operation. As another example, the state “0” at time t5 in FIG.
Assuming that there is no branch path connected to “1” or the next time point t6, the state “01” at the time point t5 is changed to the time point t2.
The branch route up to the state “10” is the route number memory unit 3
3 deleted. The number-of-paths memory unit 34 determines a surviving path finally left for the decoding depth through the above process. Therefore, if the number-of-paths memory unit 34 is designed to output the decoded data corresponding to the surviving path finally left, the decoded data corresponding to the transmitted data can be obtained. FIG. 5 shows a comparison result between the trace deletion method presented by the present invention and the decoding performance of the existing inverse tracking method for the 4-state convolutional code. In FIG. 5, a relatively thin solid line indicates a result by the method proposed by the present invention, and a relatively thick solid line indicates a result by the existing method. It can be seen from FIG. 5 that the proposed method further has a gain of about 1.2 dB at a symbol error rate of 10 ^-5 compared to the existing method . Further, the above-described trace deletion method is performed every time a state is found that is not connected to the state at the next time, or a state that is not connected to the state at the next time intermittently or after the time set previously is found. Can be done every hour.
Therefore, the determination of the point at which the trace deletion method is performed does not limit the present invention. As described above, the decoding method and apparatus using the trace elimination method in the Viterbi algorithm according to the present invention provide the n-th decoding from the (n-1) -th state.
Route number information indicating the number of branch routes that advance to the state at the time
And branch based on the generated number of paths information.
Deletion can improve decoding performance as compared with the existing scheme in which decoded data is determined through a reverse tracking process for each decoding depth.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a configuration diagram showing a conventional decoding device that uses a trace tracking method for a Viterbi algorithm. FIG. 2 is a trellis diagram for explaining decoding of data encoded by the apparatus of FIG. 1; FIG. 3 is a block diagram illustrating a decoding apparatus using a trace deletion method according to a preferred embodiment of the present invention; FIG. 4 is a flowchart for explaining a trace deletion process in the apparatus of FIG. 3; FIG. 5 is a performance comparison diagram between a conventional trace tracking method and a trace deletion method according to the present invention. [Description of Signs] 31 BMG section 32 ACS section 33 Survival memory section 34 Number of paths memory section

──────────────────────────────────────────────────続 き Continuation of the front page (72) Inventor Lee Hyung-kichi 300, Izumikawa-dong, Chang'an-gu, Suwon-si, Gyeonggi-do, Republic of Korea (56) References JP-A-62-243430 (JP, A) US Patent 5,751,374 (US, A) Ajay Dholakia, "In troduction to Conv olutional Codes wi th Application," Ma ssachusetts, USA, Kl uwer Academic Publ ishers, 1994, pages 98- 105 (58) investigated the field (Int.Cl. ^7, DB name) H03M 13 / 00-13/53

Claims (1)

(57) Claims 1. A method for decoding input data using a Viterbi algorithm, comprising: (a) a surviving path corresponding to each state at each time point based on the input data and a grid diagram; (B) searching for a surviving path having no branch path connected to the next state among the surviving paths determined in step (a), and determining the surviving path determined in step (a). Was
Reach the (n + 1) th state based on the surviving path
The detection operation of the state at the n-th point having no branch route is performed.
And (n-1) th state is detected from each state.
Indicates the number of branch routes that go to the state at the nth time point
Generating path number information ; and (c) deleting the branch path in the surviving path found in step (b) by using the path number information generated in step (b).
The method comprising repeated while dates back to a previous state to not overlap with other surviving paths branch path to be deleted has a branch path which is connected to the next state based on the final by; (d) step (c) Generating decoded data corresponding to the input data based on the surviving path left as a target. Wherein said step (c), the (c1) (n-1) th time step from the state (b) the path number information incoming branch path from the state of the detected n-th point in time (C2) determining in step (c1) whether there is a branch route that proceeds from the state in which the ( n-1 ) th branch route has been deleted to another state at the nth time; (C3) If step (c2) does not exist, step (c1) and subsequent steps are repeated while going back to the previous time. (C4) Step (c4) stops deleting the branch route if step (c2) exists. 2. The decoding method according to claim 1 , comprising: 3. An apparatus for decoding input data using a Viterbi algorithm, wherein a branch path corresponding to each state at each point in time is repeatedly determined based on the input data and a grid diagram, and a branch path is determined. First means for generating a decision vector and the number of paths information each time the decision is made, storing the decision vector generated by the first means,
Second means for outputting path information indicating each surviving path determined by the stored decision vector; storing path number information output from the first means and path information output from the second means; The operation of deleting the information of the branch route corresponding to the number-of-routes information of the first means from the stored route information corresponds to the operation of overlapping the branch route to be deleted with another surviving route having a branch route connected to the next state. not if not performed repeatedly while dates back to previously decoded, characterized in that the input data on the basis of the routing information of the finally remaining survivor path and a third means for generating a decrypted data Ji has response Device. 4. The number-of-paths information includes the number of branch paths from the state at the ( n-1 ) th time that enters the state at the nth time without having a branch path connected to the state at the ( n + 1 ) th time The decoding device according to claim 3 , wherein