KR100262303B1 - Survivor path trace back method in decoding with viterbi algorithm and apparatus thereof - Google Patents

Abstract

The present invention relates to a method and apparatus for traceback of a survival path in a decoding process using a Viterbi algorithm, and more particularly, to reduce the number of accesses of the traceback memory by introducing a separate memory of a small size storing a previous survival path. The present invention relates to a method and apparatus for backtracking a survival path. In the present invention, a separate survival path memory can be used to significantly reduce the number of accesses of the backtracking memory, thereby reducing the decoding time and reducing power consumption in terms of managing the backtracking memory.

Description

Survival path traceback method and decoding device in decoding process using Viterbi algorithm

The present invention relates to a method and apparatus for traceback of a survival path in a decoding process to which the Viterbi algorithm is applied. Particularly, the present invention provides a survivor capable of reducing the number of accesses of the traceback memory by introducing a separate memory for storing a previous survival path. A route backtracking method and apparatus are provided.

There is a need to protect data from occurrence of bit error due to communication channel environment such as weather change, nonlinear attenuation, and interference factor during data transmission in a communication system. In particular, since the mobile station communicates with low power, it is necessary to apply an encoding and decoding algorithm capable of reliable error correction. In the CDMA system, convolutional code, which is a code having excellent error correction function, is used for encoding data, and a Viterbi algorithm is adopted for reliable decoding of data.

The present invention relates to a backtracking method and a device that take a large weight in a decoding process using a Viterbi algorithm.

First, in order to explain the decoding process to which the Viterbi algorithm is applied, the Viterbi decoder will be described. Codewords received with noise are decoded by a Viterbi decoder as shown in FIG. 1 is a block diagram of a general Viterbi decoder.

Data transmitted from a demodulator is stored in a memory of the input unit 10, and the decoding unit 20 applies a Viterbi algorithm to perform actual decoding using the data. The BM unit 21 calculates Euclidean distance between the codeword of the input unit 10 and the codeword of each branch to obtain a branch metric (BM). The ACS unit 22 selects a path having a high probability and updates each state value (State Metric, SM) of the state memory (State Metric Memory, SMM) 24. The TB unit 23 stores the selection information of the ACS unit 22, and when a predetermined time elapses, the TB unit 23 traces back and decodes a path having a high probability of being equal to the original path transmitted. The decoded data is transmitted to the outside through the output unit 20.

The decoder 20 includes two memories, a state memory (SMM) 24 and a trace back memory 25 of the TB unit 23, to perform decoding by applying the Viterbi algorithm. have. These two memories store and update information about survival paths over time.

2 is a diagram for explaining the structure of a state memory and a traceback memory used in a general Viterbi decoder.

As shown in Fig. 2A, the state memory has a size of 2 ^K-1 × N. Here, K is defined as the number of registers constituting the convolutional encoder or the number of data affecting the encoding in the convolutional encoder with a cost length, and a state of 2 ^K-1 exists. N is the number of bits that do not cause an overflow when storing a state metric. The state value stored in the state memory is the cumulative value of the error used to select the most probable path.

As shown in Fig. 2B, the backtracking memory has a three-dimensional structure of 2 ^K-1 x D x 1 size. Where D is the length of the traceback memory. The traceback memory typically stores information about the survival paths for all 2 ^K-1 states at each step, over time for D time. The survival path information stored in the backtracking memory is selection information in each state in which a path having a high probability is selected.

First, the Viterbi algorithm is described. 3 is a lattice diagram for explaining the Viterbi algorithm when the constraint field K is three.

When the constraint field K is K, when there are 2K-1 states and K becomes large, it is difficult to explain the number of states rapidly increasing. Therefore, the case where K is 3 will be described as an example. When K is 3, as shown in Fig. 3, there are four states a = 00, b = 10, c = 01, and d = 11.

Branch values are calculated for every branch in the grid. Each branch represents a state transition, and each branch value is a value obtained by obtaining a Euclidean distance between the received codeword and the codeword generated according to the state transition. This process shows the similarity to the received codeword containing noise and is calculated in the BM unit 21 of the Viterbi decoder shown in FIG. In the grid, the state value is calculated and updated for every node. Each node represents a state generated when encoded by the convolutional encoder, and the state value of each point is a cumulative value obtained by adding the state value of the previous state and the branch value generated according to the state transition. In other words, each cumulative value is compared among two state transition possibilities, a small state transition is selected, and the cumulative value is updated. At this time, the minimum state having the selected information and the minimum state value for each step is transmitted to the TB unit 23. This process adds the previous state value and the branch value (Add), compares the two accumulated state values (Compare), and selects the lesser one among the accumulated state values to determine the similarity with the transmitted path. ) Is performed by the ACS unit 22 of the Viterbi decoder by performing Add-Compare-Select (ACS). The selection information and the minimum state of the survival path delivered to the TB unit 23 are used in the backtracking process to find the previous state to find the survival path, and the decoding is performed through the backtracking.

Now, the traceback of the survival path in the decoding process using the Viterbi algorithm is described. Backtracking is performed using the information of the survival path back in time to find the survival path with the highest probability as the original path sent. The backtracking process first begins by reading the survival path information using the minimum state received from the ACS unit 23 as the initial address of the backtracking memory. When the state of the previous time is obtained using the initial address, the state of the previous time becomes the address of the backtracking memory again, and if the trace back is repeated by the length of D while repeating this process, one bit is decoded as a ratio.

After the first bit is decoded, information on the survival path of each state in the D + 1 stage, which is the next stage after the first traceback D stage, is stored. At this time, the ^2K-1 × D × 1 size of the backtracking memory is assumed to be a circular queue structure, and the information on the survival paths of the states of the D + 1 stage after the D stage is the backtracking memory. Is stored at the location of step 1, which is the initial location of. Since the information of the position of the first position, which is the first position, is not necessary to decode one bit already, the new survival path information of the next stage, D + 1, is stored in the position of the first position, the first position. After the new survival path information is stored, the D length is traced back from the newly stored step to decode a new bit. Then, the new survival path information of the next step is stored, and the process of decoding the new one bit by tracing the D length back from the newly stored step is repeated.

Therefore, after the information for the first D stage is stored in the back estimation memory, decoding of each bit is started by the back tracking of the survival path, and then new survival path information is stored every hour thereafter, and the D length is stored. The decoding process is repeated one bit after decoding.

On the other hand, in the encoding process by the Viterbi algorithm, the number of transitions from any state to the next state can have two, and since the first starting point of the state is '0', for example, a convolutional encoder ( When the number of data affecting the coding in the convolutional encoder is K (when the number of states is 2 ^K-1 ), the survival path during the first K-1 stages can be known without performing backtracking. Therefore, after the step of D + (K-1), rather than D, the backtracking is started, the new survival path is saved every hour, and the backtracking of the D length is performed to decode bit by bit. Proposed.

FIG. 4 is a diagram for explaining traces of survival paths in the decoding process to which the Viterbi algorithm is applied.

As shown in Fig. 4, the first traceback starts after the survival path information for the step D + (K-1) is stored in the traceback memory. Since the length of the backtracking memory is D and has a structure of a circular queue, as shown in FIG. 4, after storing the survival path information up to the D stage, the traceback memory returns to the initial position and the D + (K− Save survival path information up to step 1). Thereafter, by using the circular queue structure, backtracking of the length D from the position K-1 is performed to decode one bit. In particular, Figure 4b shows the backtracking start time and the backtracking length of D in the backtracking memory having the structure of a circular queue.

However, in the conventional general traceback method described above, a D length traceback must be performed each time one bit is decoded, and thus the accessibility of the traceback memory is frequent and affects the total decoding time. Typically, since a previously traced survival path and a current traceback survival path are often similar, it is inefficient to perform a traceback of D length without considering these characteristics.

Accordingly, there is a need for a method and apparatus capable of efficiently decoding the access path of the backtracking memory in consideration of the characteristics that the survival path is similar to the previously tracked survival path and the current backtracked survival path.

The present invention is to solve the problems of the prior art as described above, an object of the present invention is to duplicate the survival in consideration of the characteristics that the survival path is similar in the previous trace back survival path and the current trace back survival path In the path, the present invention provides a method and apparatus for efficiently decoding the number of accesses of the backtracking memory by stopping the backtracking and using the previously tracked survival path.

1 is a block diagram of a typical Viterbi decoder;

2 is a view for explaining the structure of a state memory and a traceback memory used in a general Viterbi decoder;

3 is a lattice diagram for explaining the Viterbi algorithm when the constraint field K is 3;

FIG. 4 is a diagram for explaining the traceback of survival paths in the decoding process using the Viterbi algorithm. FIG.

5 is an exemplary diagram for explaining that two survival paths, a previous survival path and a current survival path, coincide.

Figure 6 is a block diagram of a survival path traceback device according to an embodiment of the present invention,

Figure 7 is a block diagram of a survival path traceback device according to another embodiment of the present invention.

Explanation of symbols on the main parts of the drawings

60, 70: traceback memory 61, 71: survival path memory

62, 72: traceback status register 63, 73: survival path status register

64, 74: comparators 65, 75: controller

76: parallel / serial converter 77: serial / parallel converter

In order to achieve the above object, the survival path traceback method in the decoding process to apply the Viterbi algorithm according to the present invention, the reverse path read process in the traceback with respect to the survival path information obtained by performing the initial traceback Storing the information in the tracking memory as a previous survival path; Decoding one bit as a result of the backtracking of the above step, and then storing information of one new survival path; When performing the next traceback, after performing the traceback of the new survival path stored in the step, the information of the current survival path obtained by performing the traceback in the next step and the information of the previous survival path stored in the step Comparing the; Performing a backtracking of the next step if the information of the current survival path and the information of the previous saved survival path obtained by performing the backtracking are different as a result of the comparison in the above step; As a result of the comparison in the above step, if the information of the current survival path obtained by performing the traceback and the information of the previously saved survival path are the same, if the same is stored, the survival path after the step is stored without performing the backtracking after the step. Decoding immediately using information of a previous survival path; And updating the information of the previous survival path with the information of the current survival path for the next traceback.

In order to achieve the above object of the present invention, a survival path traceback apparatus in the decoding process applying the Viterbi algorithm according to the present invention, the traceback memory for storing the survival path for each state for all states; A survival path memory for storing information of a previous survival path obtained as a result of the previous traceback; A backtracking state register for storing information of a current survival path obtained by accessing the backtracking memory step by step; A survival path state register for storing information of a previous survival path obtained by accessing the survival path memory in the same step of accessing the traceback memory; A comparator for comparing the information of the current survival path stored in the traceback status register with the information of the previous survival path stored in the survival path status register step by step; And if the information of the current survival path stored in the traceback state register and the information of the previous survival path stored in the survival path state register are the same as the result of the comparison by the comparator, after the step without performing backtracking after the step; The survival path is characterized in that it includes a controller to immediately decode by using the information of the stored previous survival path, and otherwise to continue the backtracking of the next step.

In the survival path traceback apparatus according to the present invention, when the size of the traceback memory is ^2K-1 × D, the survival path memory may be configured as a 1-bit memory having a size of 1 × D. Here, K is the number of data affecting the coding in the convolutional encoder, and the number of states is 2 ^K-1 .

In the survival path traceback apparatus according to the present invention, when the size of the traceback memory is ^2K-1 × D, the survival path memory is composed of M bit memory having the size of M × (D / M). Where M is not one.

In the survival path traceback apparatus according to the present invention, when the size of the traceback memory is ^2K-1 × D, the survival path memory is composed of M bit memory having the size of M × (D / M). If so, the survival path traceback apparatus according to the present invention comprises: a parallel / serial converter for converting parallel data output from the survival path memory into serial data input to the survival path status register; And a serial / parallel converter required for input of the survival path memory to update the previous survival path stored in the survival path memory with the current survival path.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

The backtracking memory, which stores the survival path, stores information about the status of the previous step. That is, if the current state is known and the information of the traceback memory corresponding to the state is known, the state of the previous step can be known. In addition, when performing backtracking, if any state is determined, the survival path is determined through the backtrace of D. Therefore, if the survival path backtracked in the current step and the survival path of the previous step are the same, the backtracking process in the current step can be stopped and decoded using the survival path already found in the previous step. 5 is an exemplary diagram for explaining that two survival paths coincide.

FIG. 5 shows the survival path traced back in the 20 th step t ₂₀ and the traceback survival path back in the 19 th step t ₁₉ when the constraint field K is 3 and the length of the traceback memory is 15. FIG. have. In the nineteenth step t ₁₉ , the state “0” was traced back to the minimum state, and in the twentieth step t ₂₀ , the state “11” was traced back to the minimum state. 5, the second step 15 it can be seen (t ₁₅₎ to the two survivor paths in the first match at, Therefore, in the previous step, 15 second step ₍₁₅ t) corresponds to the two survivor paths. Therefore, if you are storing the survival path of the previous step starting from the _19th step (t ₁₉ ), you can know the entire survival path through only 5 tracebacks in the backtracking of the current step starting from the _20th step (t ₂₀ ). Can be.

The present invention seeks to utilize the features as described above.

To this end, the survival path traceback apparatus in the decoding process applying the Viterbi algorithm according to the present invention newly introduces a survival path memory that stores the previous survival path obtained as a result of the previous traceback.

Figure 6 is a block diagram of a survival path traceback device according to an embodiment of the present invention.

As shown in Fig. 6, the survival path backtracking apparatus according to the present invention includes a backtracking memory 60 which stores a survival path for each state for all states, and the previous survival path obtained from the previous backtracking result. A survival path memory 61 for storing information, a backtracking status register 62 for storing information of a current survival path obtained by accessing the backtracking memory step by step, and the same for accessing the backtracking memory with the survival path memory Survival path status register 63 for storing the information of the previous survival path obtained by accessing step by step, comparing the information of the current survival path stored in the traceback status register with the information of the previous survival path stored in the survival path status register step by step Comparator 64 and a comparison result of the current survival path stored in the traceback state register as a result of the comparison by the comparator. If the information of the previous survival path stored in the beam and the survival path status register is the same, it is possible to decode immediately using the stored information of the previous survival path for the survival path after the step without performing backtracking after the step. If not, it includes a controller 65 to continue the backtracking of the next step.

The operation of the survival path backtracking device according to the present invention having such a configuration will be described. The current state survival path is traced back step by step through the traceback state register 62, and the previous state survival path stored in the survival path memory 61 is retrieved step by step through the survival path state register 63. In comparator 64 at each step, the values of both registers are compared with each other. As a result of the comparison, if the two values are the same, since they are the same as the previous survival paths, they are decoded directly to the survival path memo value of the previous step by D rather than the step of starting the backtracking. In other words, when the survival path memory information of the next step position of the current step that started backtracking is read, it is immediately decoded. However, if the values of the traceback status register 62 and the survival path status register 63 are different, the traceback memory 60 performs the following traceback through the traceback status register 62. The survival path memory 61 side has already determined the survival path, but uses the information of the survival path memory 61 in the same way as the backtracking to compare the state values at each step.

After the decoding of one bit is completed, the contents of the survival path memory 61 must be updated with the status value of the survival path of the current step before performing the backtracking of the next step.

Figure 7 is a block diagram of a survival path traceback device according to another embodiment of the present invention.

In the survival path traceback apparatus according to the present invention, when the size of the traceback memory 60 is 2 ^K-1 × D, the survival path memory may be configured as a 1-bit memory having a size of 1 × D. Can be. Here, K is the number of data affecting the coding in the convolutional encoder, and the number of states is 2 ^K-1 .

Further, in the survival path traceback apparatus according to the present invention, when the size of the traceback memory is ^2K-1 × D, the survival path memory is an M bit memory having a size of M × (D / M) (where M may not be 1), in which case, a parallel / serial converter for converting the parallel data output from the survival path memory 71 into serial data input to the survival path status register 73. And a serial / parallel converter 77 necessary for input of the survival path memory 71 to update the previous survival path stored in the survival path memory 71 with the current survival path.

Of the main parts of Viterbi decoder, the backtracking process has a large portion of power consumption in terms of power consumption. In the related art, since a traceback process having a traceback length is always required for decoding one bit after a predetermined time elapses, the traceback memory is frequently accessed.

In the present invention, a survival path memory is additionally introduced for an efficient traceback process. That is, after storing the survival path information of the previous step, when performing backtracking in the current step, a structure is used to determine whether to decode during the backtracking by comparing the backtracking state with the survival path state of the previous step. If the survival path status of the previous step and the backtracking status of the current step are the same, it means that the survival path of the previous step and the current step is the same. Therefore, the traceback process can be omitted and the information of the survival path memory can be directly output and decoded. In addition, the size of the survival path memory used is 1 × D when the size of the constraint field K is 'K', which is smaller than that of the conventional backtracking memory having a size of ^2K-1 × D. By using such a small size of the survival path memory, the number of accesses of the traceback memory can be drastically reduced. Reducing the number of accesses of the backtracking memory not only shortens the decoding time, but also reduces power consumption in terms of management of the backtracking memory.

Claims (5)

In the method for traceback of the survival path in the decoding process applying the Viterbi algorithm, Storing the information of the backtracking memory read in the backtracking process as the previous survival path with respect to the survival path information obtained by performing the initial backtracking; Decoding one bit as a result of the backtracking of the above step, and then storing information of one new survival path; When performing the next traceback, after performing the traceback of the new survival path stored in the step, the information of the current survival path obtained by performing the traceback in the next step and the information of the previous survival path stored in the step Comparing the; Performing a backtracking of the next step if the information of the current survival path and the information of the previous saved survival path obtained by performing the backtracking are different as a result of the comparison in the above step; As a result of the comparison in the above step, if the information of the current survival path obtained by performing the traceback and the information of the previously saved survival path are the same, if the same is stored, the survival path after the step is stored without performing the backtracking after the step. Decoding immediately using information of a previous survival path; And And updating the information of the previous survival path with the information of the current survival path for the backtracking of the next step. In the apparatus for tracking the survival path in the decoding process applying the Viterbi algorithm, A backtracking memory that stores survival paths for each state for all states; A survival path memory for storing information of a previous survival path obtained as a result of the previous traceback; A backtracking state register for storing information of a current survival path obtained by accessing the backtracking memory step by step; A survival path state register for storing information of a previous survival path obtained by accessing the survival path memory in the same step of accessing the traceback memory; A comparator for comparing the information of the current survival path stored in the traceback status register with the information of the previous survival path stored in the survival path status register step by step; And When the comparison result of the comparator compares the information of the current survival path stored in the traceback state register with the information of the previous survival path stored in the survival path state register, the survival after the step is performed without performing backtracking after the step. Survival in the decoding process applying the Viterbi algorithm, characterized in that it includes a controller to immediately decode by using the information of the previous survival path stored for the path, otherwise the traceback of the next step is continued. Path traceback device. 3. The method of claim 2, wherein when the size of the backtracking memory is 2 ^K-1 × D (where K is the number of data that affects coding in a convolutional encoder, the number of states is 2 ^{K−). 1} ), the survival path memory tracking device in the decoding process applying a Viterbi algorithm, characterized in that consisting of 1-bit memory having a size of 1 × D. 3. The method of claim 2, wherein when the size of the backtracking memory is 2 ^K-1 × D (where K is the number of data that affects coding in a convolutional encoder, the number of states is 2 ^{K−). 1} ), the survival path memory in the decoding process to apply the Viterbi algorithm, characterized in that the M bit memory having a size of M × (D / M) (where M is not 1). Survival path backtracking device. 5. The apparatus of claim 4, wherein the survival path traceback apparatus comprises: a parallel / serial converter for converting parallel data output from the survival path memory into serial data input to the survival path status register; And In order to update the previous survival path stored in the survival path memory to the current survival path, further includes a serial / parallel converter required for input of the survival path memory, the trace of the survival path in the decoding process applying the Viterbi algorithm Device.

Images