Satellite digital broadcast television signal physical layer scrambling parameter identification method

Abstract

The application discloses a method for identifying a scrambling code parameter of a satellite digital broadcast television signal physical layer. The application firstly constructs a Tanner graph corresponding equation, a neural network input layer, a hidden layer and an output layer, calculates a cross-entropy loss function, updates all coefficients of the network, and then adjusts the coefficients of the constructed neural network by using an arbitrary scrambling code initial value; a chip completes signal receiving, tuning, analog-digital conversion, demodulation operation, saves the obtained signal frame into a frame buffer, installs a unilateral BPSK signal solution soft information, calculates the output of the input layer according to the input layer network coefficient obtained by offline training, calculates the result of the hidden layer according to the hidden layer network coefficient obtained by offline training, calculates the result of the output layer according to the output layer network coefficient obtained by offline training, makes a hard decision on the output of the first 18 neurons of the output layer, obtains an initial value of a scrambling code sequence, and writes the initial value into a preset descrambling register of the chip. The application reduces the operation amount of offline calculation, and does not need a complex iteration process.

Description

Technical Field

[0001] This invention belongs to the field of communication technology, especially the field of digital satellite television broadcasting technology, and specifically relates to a method for identifying physical layer scrambling parameters of satellite digital broadcast television signals, which identifies the initial scrambling value of the target satellite signal source. Background Technology

[0002] In digital satellite television broadcasting applications, a cellular distribution map is typically formed by multiple satellites. Inevitably, two or more satellites will simultaneously transmit different signals on the same frequency band, covering the same area. Co-channel interference (CCI) will inevitably occur within this area; CCI is a widespread problem in existing digital communication systems. The European second-generation satellite digital television transmission standard, DVB-S2 / S2X, uses intra-frame scrambling to mitigate the sudden impact of CCI. Scrambling techniques can typically resolve both sudden and continuous narrowband CCI, working by matching the target signal while whitening out the interfering signal. This is a common measure to improve system robustness. On the other hand, for target users within the satellite coverage area who wish to simultaneously receive signals relayed from two satellites on the same frequency band, this can only be achieved by identifying the initial value of the physical layer scrambling code to distinguish the signals relayed by different satellites for multiple access communication. Therefore, identifying the physical layer scrambling code is fundamental to correctly demodulating digital satellite broadcast television signals.

[0003] The scrambling method for DVB-S2 / S2X uses a fixed-structure scrambling generator and a given initial scrambling value *n*, where *n* has a fixed range of 0 ≤ *n* ≤ 2^62141. Different initial scrambling values ​​yield different random number sequences; a specific *n* corresponds to a specific, known random sequence. The target data is scrambled using this random number sequence and then transmitted over the air interface. At the receiver, the same scrambling sequence is used to descramble the received signal, thus recovering the target signal. However, when the initial scrambling value of the target signal is unknown, the scrambling random sequence cannot be accurately obtained, making it impossible to correctly descramble the target data and recover the target data from the transmitter.

[0004] In the absence of an initial scrambling code value, the method of recovering the initial value of the random scrambling code sequence from the received interfered signal is called the scrambling code initial value identification method. In this method, since the generator polynomial is known, the simplest approach is to brute-force iterate through the relevant parameters and perform peak value judgment. However, this method requires calculating 2... 18-1 calculations is clearly difficult to achieve. Patent CN106330800B proposes a fast physical layer scrambling code parameter search method based on pilot symbols, but this method is only applicable when pilot symbols are available and cannot be used without pilots. Patent CN106330396B proposes a fast physical layer scrambling code parameter search method based on padding frames, but its drawback is the same as that of patent CN106330800B: it also fails when there are no padding frames. Patent CN103560863B proposes a pseudo-random scrambling code identification method, which identifies the initial state of the scrambling code based on convolutional correlation. Although this solves the problem of identifying the initial value of the scrambling code without known data, the storage complexity of the algorithm increases significantly due to the conversion to convolutional codes. Therefore, it suffers from high chip implementation complexity and high chip cost, ultimately making it difficult to widely apply in civilian broadcast television receivers. Summary of the Invention

[0005] The purpose of this invention is to provide a method for identifying physical layer scrambling parameters of satellite digital broadcast television signals, which solves the problem of identifying the initial scrambling value of a target satellite signal source without the assistance of pilots or empty frames.

[0006] Specifically, the present invention is:

[0007] Step (1) Construct the equation corresponding to the Tanner diagram Where F x The scrambling equation is 1+x 7 +x 18 The corresponding scrambling code state transition matrix is ​​18×18; identity matrix For mathematical symbols, This means defining a as b; v i c represents the index of the i-th variable node, 1≤i≤4626; j This represents the index of the j-th check node, 1 ≤ j ≤ 4608; for every node with a 1 in matrix H, the corresponding v is... i c j Connect them together to obtain a Tanner diagram based on matrix H;

[0008] Step (2) Construct the neural network input layer. The number of neurons in the input layer is N, where N = 4626. The input of each neuron is the soft information of the current scrambling bits, and the output of the neuron is... This refers to the output of the input layer, where l0 represents the soft information of the current scrambling bits in the input, l1 represents the neurons in the input layer, and x∈Ω(l0)\l1 represents the neurons in the input layer that are connected to the soft information in the input layer. xThe connection coefficients between the input layer neurons and the input soft information are represented by the activation function F(x) = tanh(x);

[0009] Step (3) Construct the hidden layers of the neural network. There are a total of 8 main layers in the hidden layers of the neural network. Each main layer includes two sub-layers. The first layer of all sub-layers constitutes an odd-numbered layer, and the second layer of all sub-layers constitutes an even-numbered layer.

[0010] Each sublayer consists of E neurons, corresponding to E edges in the Tanner graph. Odd-numbered layers correspond to variable nodes, and even-numbered layers correspond to check nodes. The connections between layers are determined by matrix H. The output L of each odd-numbered layer neuron... od→ev It is the variable node v i With verification node c j The side of the connecting line e = (v i ,c j The sum of the products of the data on the edge and the corresponding coefficients of the edge. Where od→ev represents the data transmitted from the odd layer to the even layer, ev'∈Ω(od)\ev represents all even-layer neurons connecting the odd layer except itself, and L od'←ev' w represents the output of an odd-numbered layer neural network. od'←ev' This represents the coefficient values ​​connecting the neural network; the output of each even-numbered layer neuron. od←ev represents data passed from even-numbered layers to odd-numbered layers;

[0011] Step (4) Construct the output layer, which consists of N neurons. The output of each neuron is... w 8,od'←ev' and L 8,od'←ev' These represent the network coefficients and output of the 8th hidden layer, respectively; the activation function σ(x) = (1 + e^(-x) / x) -x ) -1 ;

[0012] Step (5) Calculate the cross-entropy loss function Among them, u v It is the original, known training set label data, generated from a known, predefined scrambling sequence;

[0013] Step (6) Iteratively update all coefficients w in the network using gradient descent;

[0014] Step (7) Construct a set of arbitrary initial scrambling codes b = [b0, b1, ..., b 17 According to the scrambling equation F x Generate a scrambling sequence s = [s0, s1, ..., s 4607 After adding 0dB noise to the scrambling sequence, it is fed into the neural network of steps (1) to (6) to adjust the coefficient w, and the adjustment is repeated K times, K = 50 to 500;

[0015] Step (8) Reconstruct a set of arbitrary initial scrambling codes b′=[b′0,b′1,…,b′] 17 Repeat step (7) until all 218 possibilities have been traversed;

[0016] Step (9) saves the neural network coefficients w to the chip;

[0017] Step (10) The chip is powered on and completes signal reception, tuning, analog-to-digital conversion and demodulation operations. The obtained signal frame is saved to the frame buffer. The CPU retrieves the first 4608 IQ data from the frame buffer.

[0018] Step (11) Install one-sided BPSK signal soft information on 4608 IQ data points; the formula for calculating the soft information is as follows: Where P(u) n =0|y n ) and P(u n =1|y n ) represent the received IQ signals as y n In the case of , the probability that the scrambling sequence value is 0 or 1 at the current moment;

[0019] Step (12) According to the input layer network coefficients w obtained from offline training x Calculate the output of the input layer

[0020]

[0021] Step (13) Calculate the hidden layer results according to the hidden layer network coefficients obtained from offline training;

[0022] The formula for calculating odd-numbered layers is:

[0023] The formula for calculating even-numbered layers is:

[0024] Step (14) Calculate the output layer results based on the output layer network coefficients obtained from offline training. The calculation formula is as follows:

[0025]

[0026] Step (15) performs a hard decision on the outputs of the first 18 neurons of the output layer, making them either 0 or 1, such as L. out If the value is ≥0.5, it is determined as 1; otherwise, it is determined as 0. This yields the initial value a = [a0, a1, ..., a...]. 17 ];

[0027] Step (16) calculates a = [a0, a1, ..., a 17Once written into the chip's preset descrambling register, the correct TS stream decoding can be completed, and video playback can be achieved.

[0028] This invention differs from previous patents in that it utilizes known information such as empty frames and pilot signals for initial scrambling code decoding. It also avoids complex decoding algorithms using convolutional codes or low-density parity-check codes. Instead, it cleverly constructs a neural network based on the coding equation, pre-training a set of network coefficients offline and pre-setting them in a C language program. During online execution, the initial value of the scrambling sequence is calculated based on the network coefficients and real-time acquired IQ frame data. This significantly reduces the computational load of offline calculations and eliminates the need for complex iterative processes. Attached Figure Description

[0029] Figure 1 This is a schematic diagram showing the location of the method of the present invention in the system;

[0030] Figure 2 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0031] A method for identifying physical layer scrambling parameters in satellite digital broadcast television signals is applied between the demodulator and decoder. Its purpose is to calculate a correct initial scrambling value for accurate decoding by the decoder. Figure 1 As shown in the dashed box in the image.

[0032] The specific process is as follows: Figure 2 As shown:

[0033] Step (1) Construct the equation corresponding to the Tanner diagram Where F x The scrambling equation is 1+x 7 +x 18 The corresponding scrambling code state transition matrix is ​​18×18; identity matrix For mathematical symbols, This means defining a as b; v i c represents the index of the i-th variable node, 1≤i≤4626; j Let v represent the index of the j-th check node, where 1 ≤ j ≤ 4608. For each node in matrix H that has a 1, its corresponding v will be... i c j Connect them together to obtain a Tanner diagram based on matrix H.

[0034] Step (2) Construct the neural network input layer. The codeword length N = 4626. The number of neurons in the input layer is N. The input of each neuron is the soft information of the current scrambling bit. The output of each neuron is... This refers to the output of the input layer, where l0 represents the soft information of the current scrambling bits in the input, l1 represents the neurons in the input layer, and x∈Ω(l0)\l1 represents the neurons in the input layer that are connected to the soft information in the input layer. x The connection coefficients between the input layer neurons and the input soft information are represented by the activation function F(x) = tanh(x).

[0035] Step (3) Construct the hidden layers of the neural network. There are a total of 8 main layers in the hidden layers of the neural network. Each main layer includes two sub-layers. The first layer of all sub-layers constitutes an odd-numbered layer, and the second layer of all sub-layers constitutes an even-numbered layer.

[0036] Each sublayer consists of E neurons, corresponding to E edges in the Tanner graph. Odd-numbered layers correspond to variable nodes, and even-numbered layers correspond to check nodes. The connections between layers are determined by matrix H. The output L of each odd-numbered layer neuron... od→ev It is the variable node v i With verification node c j The side of the connecting line e = (v i ,c j The sum of the products of the data on the edge and the corresponding coefficients of the edge. Where od→ev represents the data transmitted from the odd layer to the even layer, ev'∈Ω(od)\ev represents all even-layer neurons connecting the odd layer except itself, and L od'←ev' w represents the output of an odd-numbered layer neural network. od'←ev' This represents the coefficient values ​​connecting the neural network; the output of each even-numbered layer neuron. od←ev represents data passed from even-numbered layers to odd-numbered layers.

[0037] Step (4) Construct the output layer, which consists of N neurons. The output of each neuron is... w 8,od'←ev' and L 8,od'←ev' These represent the network coefficients and output of the 8th hidden layer, respectively; the activation function σ(x) = (1 + e^(-x) / x) -x ) -1 .

[0038] Step (5) Calculate the cross-entropy loss function Among them, u v It is the original, known training set label data, generated from a known scrambling sequence.

[0039] Step (6) Iteratively update all coefficients w in the network using the traditional gradient descent method.

[0040] Step (7) Construct a set of arbitrary initial scrambling codes b = [b0, b1, ..., b 17 According to the scrambling equation F xGenerate a scrambling sequence s = [s0, s1, ..., s 4607 After adding 0dB noise to the scrambling sequence, it is fed into the neural network of steps (1) to (6) to adjust the coefficient w, and the adjustment is repeated K times, K = 50 to 500, and in this embodiment K = 100.

[0041] Step (8) Reconstruct a set of arbitrary initial scrambling codes b′=[b′0,b′1,…,b′] 17 Repeat step (7) until the traversal is complete. 18 One possibility.

[0042] Step (9) saves the neural network coefficients w to the chip.

[0043] The following are the steps for the chip to operate online in real time.

[0044] Step (10): The chip powers on, completes signal reception, tuning, analog-to-digital conversion, and demodulation operations, and saves the obtained signal frame to the frame buffer. The CPU retrieves the first 4608 IQ data from the frame buffer;

[0045] Step (11) installs one-sided BPSK signal soft information on 4608 IQ data points. The formula for calculating the soft information is as follows: Where P(u) n =0|y n ) and P(u n =1|y n ) represent the received IQ signals as y n In the case of , the probability that the scrambling sequence value is 0 or 1 at the current moment;

[0046] Step (12) According to the input layer network coefficients w obtained from offline training x Calculate the output of the input layer

[0047] Step (13) Calculate the hidden layer results based on the hidden layer network coefficients obtained from offline training. The calculation formula for odd-numbered layers is as follows: The formula for calculating even-numbered layers is:

[0048] Step (14) Calculate the output layer results based on the output layer network coefficients obtained from offline training. The calculation formula is as follows:

[0049]

[0050] Step (15) performs a hard decision on the outputs of the first 18 neurons of the output layer, making them either 0 or 1, such as L. out If the value is ≥0.5, it is determined as 1; otherwise, it is determined as 0. This yields the initial value a = [a0, a1, ..., a...]. 17 ].

[0051] Step (16) calculates a = [a0, a1, ..., a 17 Once written into the chip's preset descrambling register, the correct TS stream decoding can be completed, and video playback can be achieved.

Claims

1. A method for identifying physical layer scrambling parameters of satellite digital broadcast television signals, characterized in that: Step (1) Construct the equation corresponding to the Tanner diagram Where F x The scrambling equation is 1+x 7 +x 18 The corresponding scrambling code state transition matrix is ​​18×18; identity matrix For mathematical symbols, This means defining a as b; v i c represents the index of the i-th variable node, 1≤i≤4626; j This represents the index of the j-th check node, 1 ≤ j ≤ 4608; for every node with a 1 in matrix H, the corresponding v is... i c j Connect them together to obtain a Tanner diagram based on matrix H; Step (2) Construct the neural network input layer. The number of neurons in the input layer is N, where N = 4626. The input of each neuron is the soft information of the current scrambling bits, and the output of the neuron is... This refers to the output of the input layer, where l0 represents the soft information of the current scrambling bits in the input, l1 represents the neurons in the input layer, and x∈Ω(l0)\l1 represents the neurons in the input layer that are connected to the soft information in the input layer. x The connection coefficients between the input layer neurons and the input soft information are represented by the activation function F(x) = tanh(x); Step (3) Construct the hidden layers of the neural network. There are a total of 8 main layers in the hidden layers of the neural network. Each main layer includes two sub-layers. The first layer of all sub-layers constitutes an odd-numbered layer, and the second layer of all sub-layers constitutes an even-numbered layer. Each sublayer consists of E neurons, corresponding to E edges in the Tanner graph. Odd-numbered layers correspond to variable nodes, and even-numbered layers correspond to check nodes. The connections between layers are determined by matrix H. The output L of each odd-numbered layer neuron... od→ev It is the variable node v i With verification node c j The side of the connecting line e = (v i ,c j The sum of the products of the data on the edge and the corresponding coefficients of the edge. Where od→ev represents the data transmitted from the odd layer to the even layer, ev'∈Ω(od)\ev represents all even-layer neurons connecting the odd layer except itself, and L od'←ev' w represents the output of an odd-numbered layer neural network. od'←ev' This represents the coefficient values ​​connecting the neural network; the output of each even-numbered layer neuron. od←ev represents data passed from even-numbered layers to odd-numbered layers; Step (4) Construct the output layer, which consists of N neurons. The output of each neuron is... w 8,od'←ev' and L 8,od'←ev' These represent the network coefficients and output of the 8th hidden layer, respectively; the activation function σ(x) = (1 + e^(-x) / x) -x ) -1 ; Step (5) Calculate the cross-entropy loss function Among them, u v It is the original, known training set label data, generated from a known, predefined scrambling sequence; Step (6) Iteratively update all coefficients w in the network using gradient descent; Step (7) Construct a set of arbitrary initial scrambling codes b = [b0, b1, ..., b 17 According to the scrambling equation F x Generate a scrambling sequence s = [s0, s1, ..., s 4607 After adding 0dB noise to the scrambling sequence, it is fed into the neural network of steps (1) to (6) to adjust the coefficient w, and the adjustment is repeated K times. Step (8) Reconstruct an arbitrary set of initial scrambling codes b′=[b′0,b′1,…,b′] 17 Repeat step (7) until all 218 possibilities have been traversed; Step (9) saves the neural network coefficients w to the chip; Step (10) The chip is powered on and completes signal reception, tuning, analog-to-digital conversion and demodulation operations. The obtained signal frame is saved to the frame buffer. The CPU retrieves the first 4608 IQ data from the frame buffer. Step (11) Install one-sided BPSK signal soft information on 4608 IQ data points; the formula for calculating the soft information is as follows: Where P(u) n =0|y n ) and P(u n =1|y n ) represent the received IQ signals as y n In the case of , the probability that the scrambling sequence value is 0 or 1 at the current moment; Step (12) According to the input layer network coefficients w obtained from offline training x Calculate the output of the input layer Step (13) Calculate the hidden layer results according to the hidden layer network coefficients obtained from offline training; The formula for calculating odd-numbered layers is: The formula for calculating even-numbered layers is: Step (14) Calculate the output layer results based on the output layer network coefficients obtained from offline training. The calculation formula is as follows: Step (15) performs a hard decision on the outputs of the first 18 neurons of the output layer, making them either 0 or 1, such as L. out If the value is ≥0.5, it is determined as 1; otherwise, it is determined as 0. This yields the initial value a = [a0, a1, ..., a...]. 17 ]; Step (16) calculates a = [a0, a1, ..., a 17 Once written into the chip's preset descrambling register, the correct TS stream decoding can be completed, and video playback can be achieved.

2. The method for identifying physical layer scrambling parameters of satellite digital broadcast television signals as described in claim 1, characterized in that: In step (7), K = 50 to 500.

Images