KR20020051549A - Quantization and metric estimation method for iterative decoder - Google Patents

Abstract

PURPOSE: A quantization method and a metric calculation method in a repetitive decoder are provided, which quantize a signal received by a BPSK or QPSK modulation into a digital signal of q bits and calculate the metric using the above value. CONSTITUTION: According to the quantization method and the metric calculation method in a repetitive decoder of a digital communication system receiver correcting an error generated in a channel by receiving a BPSK- or QPSK-modulated signal through the channel, the method includes the first step of quantizing the above received signal into bits of an integer(q) above at least 2, and the second step of performing a soft judgement output viterbi decoding repetitively using the quantized received signal. According to the first step, the received signal is compared with judgement levels(qlevel(ql)) which are tabled in advance, and then is quantized into the above q bits according to the judgement level.

Description

Quantization and metric estimation method for iterative decoder

The present invention relates to a quantization method and a metric calculation method used when implementing an iterative decoding algorithm using a soft decision output Viterbi algorithm in hardware.

In general, an iterative decoding method using a soft decision output (generally referred to as a turbo code) includes a method of using convolutional code and a method of using block code, which require calculation of a real value. However, in order to implement real hardware, since a real value operation cannot be performed, a technique for efficiently quantizing them and a metric calculation method using the quantized values are required.

The method of using the convolutional code is disclosed in US Pat. No. 5,446,747, which is a MAP algorithm of trellis decoding technique (LR Bahl, J. Cocke, F. Jelinek, and J. Raviv, "Optimal decoding of linear codes for minimizing symbol error rate ", IEEE Transactions on Information Theory, Vol. 20, pp. 284-287, Mar. 1974.). It is not mentioned.

A method of using a block code is disclosed in US Pat. No. 5,563,897, which uses an algebraic decoding method, which is also not mentioned here.

A method of reducing complexity and repeatedly performing decoding by applying a soft decision output Viterbi decoding method to a block code is disclosed in Korean Patent Application No. 99-35893, which is applied to a block turbo code connected in parallel. In a very efficient way. However, as described above, in the iterative decoding method using the soft decision output, only the principle of decoding is mentioned, but not the quantization method and the metric calculation method using the same.

Therefore, there is a need for a quantization method and a metric calculation method using the quantization value for implementing turbo code in real hardware.

An object of the present invention devised to meet the needs of the prior art as described above, in performing iterative decoding using a soft decision output Viterbi decoding method, the signal received by BPSK or QPSK modulated to a digital signal consisting of q bits To provide a method of quantizing and using this value to calculate the metric and normalizing the soft decision output.

1 is a block diagram of a digital communication system including a repeating decoder to which the present invention is applied;

2 is an operation flowchart showing a received signal determination level determination algorithm according to the present invention;

3 is an operation flowchart showing a quantization value determination algorithm according to the present invention;

4 is an operation flowchart showing a soft decision output Viterbi decoding algorithm according to the present invention;

FIG. 5 is a graph of bit error rate performance after six iterations when real number operation is used for a serial concatenated block turbo code using a (31,25) expurgated BCH code as an internal component code and when a quantization metric is used. FIG. .

※ Explanation of code about main part of drawing ※

11: source information 12: source encoder

13 Channel Encoder 14 BPSK / QPSK Modulator

15 quantizer 16: channel decoder

17: source decoder 18: information sink

According to the present invention for achieving the above object, a quantization and metric calculation method in a repetitive decoder of a digital communication system receiver receiving a BPSK or QPSK modulated signal through a channel and correcting an error occurring in the channel,

A first step of quantizing the received signal into at least two integer (q) bits;

And a second step of repeatedly performing soft decision output Viterbi decoding by using the quantized received signal.

Hereinafter, the "quantization and metric calculation method in a repeating decoder" according to an embodiment of the present invention with reference to the accompanying drawings will be described in more detail.

1 is a block diagram of a digital communication system including a repeater decoder to which the present invention is applied. These are output from the source encoder 12 for receiving the analog source information 11 and encoding them into a digital signal, the channel encoder 13 for channel encoding the digitized source signal output from the source encoder 12, and the channel encoder 13. BPSK / QPSK modulator 14 for transmitting QPSK-modulated coded digital information through a channel, quantizer 15 for receiving QPSK-modulated signals through a channel, and outputting soft decision information. The channel decoder 16 decodes the quantized soft decision information outputted from the CDMA, the source decoder 17 reconstructing the decoded information, and the information sink 18 extracting information from the reconstructed analog signal.

The present invention is applied to the quantization method of the quantizer 15 and the metric calculation method of the channel decoder 16. The input of quantizer 15 is an analog signal plus a signal transmitted from the transmitter plus an error occurring in the channel. The quantizer 15 assigns the received signal to the quantized soft decision value by the algorithm shown in FIG. 3 on the basis of the received signal determination level previously tabled by the algorithm as shown in FIG.

The channel decoder 16 receiving the quantized soft decision value performs the soft decision output Viterbi algorithm as shown in FIG. 4 using the branch metric calculation method shown in Equation 1 below. During decoding, the maximum value is always kept at (2 ^q -1) so that the soft decision output value is not greater than (2 ^q -1) (Note: q is an exponent). When decoding is completed, the soft decision output value calculates the external input information to be used for the next decoding by subtracting the external input information and the channel reliability value used for the current decoding. Also, the calculated external input information is multiplied by 0.25 to the external input information for normalization. This is done by shifting the whole value to the right twice.

A process of determining the reception signal determination level will be described with reference to FIG. 2. The quantizer initializes the variables i and ql, and quantizes the received signal r with q bits by comparing the received signals r with the previously determined ql decision levels (S22, S23, S24) (S25 through S29). In this case, it is assumed that the reception signal r is transmitted by normalizing '0' bit to '-1' and '1' bit to '1' at the transmitter.

Therefore, when a 3-bit quantized value is assigned to the received signal, the relationship with the decision level qlevel [i] for the level ql of the received signal is as follows.

ql = 7,

qlevel [0] = 6/7 ≒ 0.857143,

qlevel [1] = 4/7 ≒ 0.571429,

qlevel [2] = 2/7 ≒ 0.285714,

qlevel [3] = 0,

qlevel [4] = -2/7 ≒ -0.285714,

qlevel [5] = -4/7 ≒ -0.571429,

qlevel [6] = -6/7 ≒ -0.857143.

The relationship between the quantization value and the determination level qlevel [i] when allocating the three-bit quantized value is summarized in Table 1 below.

Range of judgment level to which received signal r belongs Assigned quantization value (the value of {} represents the sign bit) r <qlevel [6] {-} 11 qlevel [6] <r <qlevel [5] {-} 10 qlevel [5] <r <qlevel [4] {-} 01 qlevel [4] <r <qlevel [3] {-} 00 qlevel [3] <r <qlevel [2] {+} 00 qlevel [2] <r <qlevel [1] {+} 01 qlevel [1] <r <qlevel [0] {+} 10 qlevel [0] <r {+} 11

The quantizer according to the present invention determines the quantization value using Table 1 above. FIG. 3 is an operation flowchart illustrating this quantization value determination algorithm.

First, the variable t is initialized (S31), and it is determined whether the received signal is greater than the t-th determination level (S32). If the result of the determination in step S32 is no, the value of t is increased by one (S33), and it is determined whether t is a value smaller than ql (S34). If the determination result of step S34 is YES, the flow returns to step S32. If the determination result of step S34 is NO, the flow advances to step S35 to confirm the-(ql + 1) / 2 value as a quantization value and ends (S35). ).

On the other hand, if the determination result in step S32 is YES, it is determined whether the t-th determination level is greater than zero (S36). If the result of the determination in step S36 is YES, (ql + 1) / 2-t is determined as the quantization value and is terminated (S37). If the result of the determination in step S36 is no, (ql + 1) / 2- (t + 1) is determined as the quantization value and ends (S38).

Next, a metric calculation algorithm in the channel decoder will be described with reference to FIG. 4. The channel reliability L (u) is calculated from the received signal (S41), and the calculated channel reliability information is quantized to q bits. Quantization of this channel reliability information uses the quantization value determination method of FIG. Next, all initial external input information for the current code is initialized to 0 (S43), and the soft decision and hard decision output are calculated using the soft decision output Viterbi algorithm for the current sign using the external input information (S44). ).

If it satisfies the condition of repetition stop, if it is satisfied, it outputs hard decision decoded value and ends (S45, S49). Otherwise, it calculates external information to be used for the next code from the soft decision output for the current code. (S46). Next, two bits are shifted to the right in order to normalize the external information, and the external input information is multiplied by 0.25 (S47). Then, the code is moved to the code to be decoded and repeated from step S44 (S48).

In the soft decision output Viterbi algorithm in step S44, the branch metric (bm) is obtained by Equation 1 below, and the path metric is the branch currently calculated on the path metric at the previous point in time. Add metric to get

'

Here, r _q 'and e _q ' represent the quantized value e _q of the external input information obtained from the quantized received signal r _q and the soft decision output decoding of the previous stage, respectively, as an absolute value and a sign bit. Is determined as follows. In other words, if the sign of r _q or e _q is equal to the sign of the sign bit on the trellis at the time, the sign is positive (+) and negative if the sign is different from each other.

5 is a simulation performance diagram of a serial concatenated block turbo code using (31,25) expurgated BCH code as an internal component code using the method proposed in the present invention. In Fig. 5, the bit error rate after six iterations is compared, and it can be seen that the quantization and metric calculation methods proposed in the present invention exhibit almost the same performance as the real number operation even when using a 5-bit quantized value.

While the invention has been described above based on the preferred embodiments thereof, these embodiments are intended to illustrate rather than limit the invention. It will be apparent to those skilled in the art that various changes, modifications, or adjustments to the above embodiments can be made without departing from the spirit of the invention. Therefore, the protection scope of the present invention will be limited only by the appended claims, and should be construed as including all such changes, modifications or adjustments.

As described above, according to the present invention, in the iterative decoding method using the soft decision output Viterbi algorithm, not only the quantization method and the metric calculation method can be easily implemented in hardware, but also the number of quantization bits, i. Even if you use it, you can get almost the same performance as real operation.

Claims (8)

A quantization and metric calculation method in an iterative decoder of a digital communication system receiver which receives a BPSK or QPSK modulated signal through a channel and corrects an error occurring in the channel. A first step of quantizing the received signal into at least two integer (q) bits; And a second step of repeatedly performing soft decision output Viterbi decoding using the quantized received signal. The method of claim 1, The first step is, Comparing the received signal with 2 ^q -1 decision levels (qlevel [ql], where an integer of 0≤ql <2 ^q -1), which is pre-tabled, and quantizing the ^q- signs into q bits according to the decision level. A quantization and metric calculation method in a repeating decoder characterized by the above-mentioned. The method of claim 2, The determination level qlevel [i] (i is any integer) according to the received signal is If 0 ≤ i <ql / 2 then ((ql / 2-i) * 2) / ql, 0 if i = (ql + 1) / 2 A quantization and metric calculation method in an iterative decoder, wherein ql / 2 <i <ql, -1 * qlevel [ql-1-i]. The method of claim 3, wherein The quantization value according to the received signal is, By increasing t from 0 to 1, the t judgment level is compared with the received signal. If the received signal is greater than a t th judgment level (qlevel [t], t < ql), and an arbitrary judgment level (qlevel [t]) is greater than zero, the quantization value is (ql + 1) / 2-t; If the received signal is greater than the t th judgment level (qlevel [t], t <ql) and the arbitrary judgment level (qlevel [t]) is not greater than zero, the quantization value is (ql + 1) / 2- (t + 1) is; If the received signal does not find a value larger than the t th judgment level until t is ql, the quantization value is -1 * (ql + 1) / 2 quantization and metric calculation method in the iterative decoder. The method of claim 1, The second step, A first sub-step of determining and quantizing channel reliability using the received signal; A second substep of initializing initial external information and calculating soft decision and hard decision output using a soft decision output Viterbi algorithm for a current code using the initial external information, A third substep of outputting a hard decision decoding value if the repetition stop condition is satisfied and ending decoding, and calculating external information to be used for the next code from the soft decision output for the current code if the repetition stop condition is not satisfied; and And a fourth sub-step of normalizing the external information and moving to a code to be subjected to the next decoding. The method of claim 5, wherein the external information normalization step of the fourth sub-step includes: The quantization and metric calculation method in an iterative decoder characterized in that shifting the external information two bits to the right. The method of claim 5, The soft decision output Viterbi algorithm of the second sub-stage uses a branch metric, and the path metric uses a path metric from a previous time point and a current branch metric. Quantization and Metric Computation in. The method of claim 7, wherein The branch metric (bm) is a quantization and metric calculation method in the iterative decoder, characterized in that calculated by the following formula. here, and It will display the absolute value and the sign bit for the value e _q quantization of the external input information obtained from the soft decision output decoding of a received signal r _q from the previous stage, each quantizer and each code bit of the r _q or e _q code as At this point, if the sine of the sign bit on the trellis is the same, it is positive (+), and if they are different from each other, it is determined as a negative (-1) value.