LSTM-based laser chaotic synchronization communication system
By using LSTM-based neural network modeling, the problem of difficult laser parameter matching at the receiver end in chaotic optical communication was solved, and efficient, convenient and secure synchronous communication was achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-03-08
- Publication Date
- 2026-04-03
AI Technical Summary
Existing chaotic optical communication systems require a high degree of matching between the laser parameters of the transmitter and receiver, making efficient synchronization difficult and resulting in high system complexity.
The receiver of the chaotic optical communication system is modeled using an LSTM-based neural network. By utilizing the nonlinear fitting capability of LSTM, high-quality synchronous communication is achieved through training the neural network.
It reduces time loss, solves the problem of difficult matching of hardware parameters between the receiving and receiving parties, and realizes convenient and high-quality synchronous communication.
Smart Images
Figure CN116455472B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of chaotic laser communication technology, specifically relating to a laser chaotic synchronous communication system based on LSTM. Background Technology
[0002] Laser chaotic communication systems, with their unique advantages such as strong randomness, noise-like characteristics, and high bandwidth, are widely used in secure communication, enabling real-time synchronization and high-speed data extraction. Currently, laser chaotic synchronous communication generally relies on the nonlinear effects within the laser or photoelectric oscillators. However, existing chaotic synchronous optical communication systems require a high degree of matching between the parameters of the lasers and other devices at the transmitting and receiving ends. This ideal situation is difficult to achieve in reality. Therefore, existing chaotic optical communication systems face technical challenges such as complex hardware systems and low synchronization coefficients. Summary of the Invention
[0003] To address the shortcomings of the existing technologies, this invention provides a laser chaotic synchronous communication system based on LSTM. This invention utilizes the powerful nonlinear fitting capability of neural networks to model the receiver of the chaotic optical communication system, thereby achieving high-quality synchronous communication.
[0004] To achieve the above objectives, the present invention adopts the following technical solution:
[0005] A laser chaotic synchronization communication system based on LSTM includes a reflector (1), a first semiconductor laser (2-1), an optical fiber isolator (3), a first variable optical attenuator (4-1), a first optical coupler (5-1), a first photodiode (6-1), a second semiconductor laser (2-2), a Mach-Zehnder modulator (7), an arbitrary waveform generator (8), a second variable optical attenuator (4-2), a second optical coupler (5-2), an optical fiber (9), a second photodiode (6-2), an analog-to-digital converter (10), an LSTM network module (11), and a subtractor (12). The input port of the first semiconductor laser (2-1) faces the reflector (1), and the output port of the first semiconductor laser (2-1) is connected to the first port of the first optical coupler (5-1) in sequence through the optical fiber isolator (3) and the first variable optical attenuator (4-1). The second port of the first optical coupler (5-1) is connected to the second optical coupler (5-2). The second port of the first optical coupler (5-1) is connected to the second port of the first photodiode (6-1); the second semiconductor laser (2-2) and the arbitrary waveform generator (8) are both connected to the Mach-Zehnder modulator (7). The Mach-Zehnder modulator (7) is connected to the first port of the second optical coupler (5-2) through the second variable optical attenuator (4-2). The third port of the second optical coupler (5-2) is connected to the first port of the second photodiode (6-2) through the optical fiber (9). The second port of the second photodiode (6-2) is connected to the first port of the analog-to-digital converter (10). The third port of the analog-to-digital converter (10) is connected to the first port of the LSTM network module (11). The second port of the analog-to-digital converter is connected to the first port of the subtractor (12). The second port of the LSTM network module (11) is connected to the third port of the subtractor (12). The second port of the subtractor (12) outputs the decryption information.
[0006] The communication steps of this invention are as follows:
[0007] S1: The laser chaotic system generates a chaotic carrier signal;
[0008] S2: The information system generates information signals;
[0009] S3: The chaotic carrier signal generated by S1 and the information signal generated by S2 are coupled in a certain proportion to form an encrypted signal;
[0010] S4: The encrypted signal generated in step S3 and the chaotic carrier signal generated in step S1 are used as the input variable and target output variable of the LSTM (Long Short-term Memory) neural network, respectively, and the neural network is trained.
[0011] S5: The information signal is obtained by subtracting the predicted carrier signal from the encrypted signal and decrypting it.
[0012] As a preferred embodiment, the first semiconductor laser 2-1 with a wavelength of 1550nm outputs continuous light, which is fed back to the first semiconductor laser through an external reflector and perturbed to generate a chaotic carrier signal.
[0013] As a preferred embodiment, the information light is generated by a second semiconductor laser with a wavelength of 1550nm, which is modulated by a Mach-Zehnder modulator driven by a non-return-to-zero (NRZ) code signal, and the random NRZ signal is generated by an arbitrary waveform generator.
[0014] As a preferred embodiment, the chaotic signal and the information signal are mixed through a second optical fiber coupler to form an encrypted signal, and the mixing ratio of the two is controlled by a first optical attenuator and a second optical attenuator.
[0015] Meanwhile, considering that there will be some link damage during channel transmission, 30dB of Gaussian noise is added to the encrypted signal.
[0016] Meanwhile, in order to ensure system security and achieve better encryption, the mixing ratio of chaotic carrier signal and information signal is set to 0.08.
[0017] As a preferred approach, during the offline training phase, the encrypted signal and the chaotic carrier signal are used as the input and target output variables of the neural network, respectively. Each iteration of the neural network updates the state of its network nodes until the ideal loss value is reached. The better the training effect, the higher the fit of the neural network to the receiver, and the higher the accuracy between the predicted chaotic carrier and the target carrier.
[0018] As a preferred solution, after the encrypted signal is received by the second photodiode at the receiving end and converted into an electrical signal, the analog-to-digital converter converts the analog signal into a digital signal. The signal then passes through a pre-trained LSTM neural network to generate a chaotic carrier that is highly fitted to the transmitting end. Finally, the received encrypted information is subtracted from the synchronized chaotic carrier to decrypt the useful information.
[0019] Compared with existing technologies, the advantages of this invention are:
[0020] 1. In the technical solution of this invention, LSTM, with its characteristics of learning long-term dependencies in time series and strong robustness, greatly reduces time loss and saves the time cost of training neural networks.
[0021] 2. The technical solution proposed in this invention achieves high-quality synchronous communication in an optical feedback system, successfully solving the problem of difficult matching of hardware parameters between the receiving and receiving parties in traditional chaotic optical communication. It also has the advantages of convenience and security, providing ideas for subsequent research on chaotic optical communication. Attached Figure Description
[0022] Figure 1 This is a schematic diagram of the architecture of an LSTM-based optical feedback communication system according to the present invention.
[0023] Figure 2 This is a diagram showing the information sequence of the transmitter in the optical feedback system of the present invention. Wherein: (a) chaotic carrier sequence; (b) original information sequence; (c) noisy information sequence; (d) encrypted signal.
[0024] Figure 3 This is a decryption information diagram of the receiver of the optical feedback system of the present invention. (a) Synchronization scatter plot of the target chaotic sequence and the predicted chaotic sequence; (b) Decryption information diagram; (c) Quantization decryption information diagram. Detailed Implementation
[0025] To better illustrate the technical solution of the present invention, the specific embodiments of the present invention will be described below in conjunction with the accompanying drawings.
[0026] like Figure 1 As shown, this embodiment of the LSTM-based laser chaotic synchronous communication system includes a transmitter and a receiver connected together. Specific components include: a reflector 1, a first semiconductor laser 2-1, an optical fiber isolator 3, a first variable optical attenuator 4-1, a first optical coupler 5-1, a first photodiode 6-1, a second semiconductor laser 2-2, a Mach-Zehnder modulator 7, an arbitrary waveform generator 8, a second variable optical attenuator 4-2, a second optical coupler 5-2, an optical fiber 9, a second photodiode 6-2, an analog-to-digital converter 10, an LSTM network module 11, and a subtractor 12.
[0027] The specific connection methods for the above components are as follows:
[0028] The input port b1 of the first semiconductor laser 2-1 illuminates the port a1 of the reflector 1. The output port b2 of the first semiconductor laser 2-1 is connected to the first port c1 of the fiber optic isolator 3. The second port c2 of the fiber optic isolator 3 is connected to the first port d1 of the first variable optical attenuator 4-1. The second port d2 of the first variable optical attenuator 4-1 is connected to the first port e1 of the first optical coupler 5-1. The second port e2 of the first optical coupler 5-1 is connected to the second port g2 of the second optical coupler 5-2. The third port e3 of the first optical coupler 5-1 is connected to the first port f1 of the first photodiode 6-1. The second port f2 of the first photodiode 6-1 outputs a chaotic carrier signal, which serves as the target output of the LSTM neural network for offline training of the LSTM neural network.
[0029] The first port h1 of the second semiconductor laser 2-2 is connected to the first port j1 of the Mach-Zehnder modulator 7. The first port k1 of the arbitrary waveform generator 8 is connected to the second port j2 of the Mach-Zehnder modulator 7. The third port j3 of the Mach-Zehnder modulator 7 is connected to the first port m1 of the second variable optical attenuator 4-2. The second port m2 of the second optical attenuator 4-2 is connected to the first port g1 of the second optical coupler 5-2. The third port g3 of the second optical coupler 5-2 is connected to the first port n1 of the optical fiber 9. The second port n2 of the optical fiber 9 is connected to the first port p1 of the second photodiode 6-2. The second port p2 of the second photodiode 6-2 is connected to the first port q1 of the analog-to-digital converter 10. The third port q3 of the analog-to-digital converter 10 is connected to the first port r1 of the LSTM network module 11. The second port q2 of the analog-to-digital converter is connected to the first port s1 of the subtractor 12. The second port r2 of the LSTM network module 11 is connected to the third port s3 of the subtractor 12. The second port s2 of the subtractor 12 outputs the decrypted information.
[0030] like Figure 1 The working principle of the chaotic synchronization method based on optical feedback system is as follows:
[0031] The first semiconductor laser 2-1 outputs continuous light with a wavelength of 1550nm. This light is fed back to the first semiconductor laser 2-1 by an external reflector 1, which perturbs it to generate a chaotic carrier signal. The output light then passes through an optical fiber isolator 3 and a first variable optical attenuator 4-1, and is split into two beams by a 50:50 first optical coupler 5-1. One beam is received by a first photodiode 6-1, and this beam is used for offline training data at the neural network output. The other beam is mixed with the information light through a second optical fiber coupler 5-2 to form the original encrypted signal. This information light is generated by a Mach-Zehnder modulator 7 driven by a non-return-to-zero (NRZ) code signal, modulating the second semiconductor laser 2-2 with a wavelength of 1550nm. The random NRZ code signal is generated by an arbitrary waveform generator 8. The chaotic signal and the information signal are mixed through the second optical fiber coupler 5-2 to form the encrypted signal. The mixing ratio of the chaotic signal and the information signal is controlled by the first optical attenuator 4-1 and the second optical attenuator 4-2. To ensure system security and achieve good encryption, the mixing ratio of the chaotic carrier signal and the information signal is set to 0.08 in this embodiment. The original encrypted signal is affected by noise during channel transmission, eventually forming an encrypted signal (which includes noise). At the receiving end, the second photodiode 6-2 receives the encrypted signal and converts it into an electrical signal. The analog-to-digital converter 10 then converts the analog signal into a digital signal, which is then passed through a pre-trained LSTM (Long Short-Term Memory) neural network module 11 to generate a chaotic carrier that closely matches the transmitting end. Subsequently, the received encrypted information is subtracted from the synchronized chaotic carrier to decrypt the useful information.
[0032] During the offline training phase, the hyperparameters of the neural network structure in the LSTM network module (11) include 200 LSTM nodes, a dropout layer with a dropout rate of 0.5, and a fully connected layer. The training optimizer used is adam, the maximum number of training epochs is 250, the initial learning rate is 0.005, and the learning rate is multiplied by a decrease factor of 0.2 every 125 epochs. The encrypted signal and the chaotic carrier signal are used as the input variable and the target output variable of the neural network, respectively. Each iteration of the neural network updates the state of its network nodes until the ideal loss value is reached. The better the training effect, the higher the fit of the neural network to the receiver, and the higher the accuracy between the predicted chaotic carrier and the target carrier.
[0033] In this embodiment, considering that there will be a certain degree of link damage during channel transmission, 30dB Gaussian noise is added to the information signal at the transmitting end using an arbitrary waveform generator 8.
[0034] See Figure 2 , Figure 3In one specific embodiment, the technical solution of the present invention analyzes the communication effect between the system transmitter and receiver.
[0035] See Figure 2 For the transmitting end, the chaotic signal generated by the optical feedback semiconductor laser system serves as the chaotic carrier sequence, while a binary non-return-to-zero signal with a signal-to-noise ratio of 30dB is used as the noisy information. The two are mixed at a ratio of 0.08 to form the chaotic encryption signal. Observation of the chaotic sequence and the encrypted information shows that the amplitude difference between the two is extremely small, making them indistinguishable. This indicates that the information with a mixing ratio of 0.08 is effectively coupled into the chaotic sequence.
[0036] See Figure 3 The neural network structure comprises 200 LSTM nodes, a dropout layer with a dropout rate of 0.5, and a fully connected layer, using the Adam training optimizer. The system ultimately achieves chaotic synchronization communication of encrypted signals with a coupling coefficient of 0.08, a signal-to-noise ratio of 30dB, and a transmission rate of 2Gbit / s. After modeling and training the laser chaotic system using the LSTM neural network, the receiver's chaotic synchronization and decryption are as follows... Figure 3 As shown. Figure 3 In (a), the horizontal and vertical axes represent the actual target chaotic sequence and the chaotic sequence predicted by LSTM, respectively. The two are highly similar, with most of them fitting to the y=x line. The calculated RMSE is 8.866×10⁻⁶. -3 The system synchronization coefficient is as high as 0.9997, indicating that the system has good prediction performance and high-quality chaotic synchronization.
[0037] The encrypted information obtained by the receiving end is directly subtracted from the noisy information demodulated by the chaotic carrier predicted by the neural network, such as... Figure 3 As shown in (b), the signal is distributed around 1 and 0. At this point, the decrypted information contains slight noise, and the bit error rate (BER) calculated using the Q factor is 1.16 × 10⁻⁶. 10 As can be seen, the BER is far below the Hard Decision Threshold of Forward Error Correlation (HDFEC) value of 3.8 × 10⁻⁶. -3 This indicates that the system possesses high-quality chaotic synchronous communication capabilities. The information after quantitative demodulation is as follows: Figure 3 As shown in (c), it is consistent with the original information. Figure 2 (b) Almost identical, with higher system communication quality.
[0038] In summary, the LSTM-based laser chaotic synchronous communication scheme proposed in this invention successfully solves the problem of difficult hardware parameter matching between the receiving and receiving parties in traditional chaotic optical communication. It also has the advantages of convenience and security, and provides ideas for subsequent research on chaotic optical communication.
[0039] This invention discloses a laser chaotic synchronous communication system based on LSTM. Utilizing the powerful nonlinear fitting capability of neural networks, the receiver of the chaotic optical communication system is modeled to achieve high-quality synchronous communication. In this system: a laser chaotic system generates a chaotic carrier signal; an information system generates an information signal; the chaotic carrier signal and the information signal are coupled at a certain ratio to form an encrypted signal; the encrypted signal and the chaotic carrier signal serve as the input and target output variables of the LSTM neural network, respectively, and the neural network is trained; the information signal is obtained by subtracting the predicted carrier signal from the encrypted signal. The LSTM-based laser chaotic synchronous communication scheme proposed in this invention successfully solves the problem of difficult hardware parameter matching between the receiving and receiving parties in traditional chaotic optical communication, while also possessing high convenience and security.
[0040] The accompanying drawings described above are merely some embodiments of the present invention. For those skilled in the art, other drawings and other implementation methods can be obtained based on these drawings without any creative effort.
Claims
1. A laser chaotic synchronization communication system based on LSTM, characterized in that, The system includes a reflector (1), a first semiconductor laser (2-1), an optical fiber isolator (3), a first variable optical attenuator (4-1), a first optical coupler (5-1), a first photodiode (6-1), a second semiconductor laser (2-2), a Mach-Zehnder modulator (7), an arbitrary waveform generator (8), a second variable optical attenuator (4-2), a second optical coupler (5-2), an optical fiber (9), a second photodiode (6-2), an analog-to-digital converter (10), an LSTM network module (11), and a subtractor (12). The input port of the first semiconductor laser (2-1) faces the reflector (1). The output port of the first semiconductor laser (2-1) is connected to the first port of the first optical coupler (5-1) in sequence through the fiber optic isolator (3) and the first variable optical attenuator (4-1). The second port of the first optical coupler (5-1) is connected to the second port of the second optical coupler (5-2). The third port of the first optical coupler (5-1) is connected to the first port of the first photodiode (6-1). The second semiconductor laser (2-2) and the arbitrary waveform generator (8) are both connected to the Mach-Zehnder modulator (7). The Mach-Zehnder modulator (7) is connected to the first port of the second optical coupler (5-2) through the second variable optical attenuator (4-2). The third port of the second optical coupler (5-2) is connected to the first port of the second photodiode (6-2) through the optical fiber (9). The second port of the second photodiode (6-2) is connected to the first port of the analog-to-digital converter (10). The third port of the analog-to-digital converter (10) is connected to the first port of the LSTM network module (11). The second port of the analog-to-digital converter is connected to the first port of the subtractor (12). The second port of the LSTM network module (11) is connected to the third port of the subtractor (12). The second port of the subtractor (12) outputs the decryption information.
2. The LSTM-based laser chaotic synchronization communication system according to claim 1, characterized in that: The first semiconductor laser (2-1) outputs continuous light with a wavelength of 1550nm, which shines on the reflector. The reflector feeds back to the input port of the first semiconductor laser and perturbs it to generate a chaotic carrier signal.
3. The LSTM-based laser chaotic synchronization communication system according to claim 2, characterized in that: The Mach-Zendel modulator (7) driven by the non-return-to-zero (NRZ) signal modulates the second semiconductor laser (2-2) with a wavelength of 1550 nm to generate an information optical signal, and the random NRZ signal is generated by the arbitrary waveform generator (8).
4. The LSTM-based laser chaotic synchronization communication system according to claim 3, characterized in that: The chaotic carrier signal and the information optical signal are mixed through the second optical fiber coupler (5-2) to form an encrypted signal. The mixing ratio of the two is controlled by the first optical attenuator (4-1) and the second optical attenuator (4-2).
5. The LSTM-based laser chaotic synchronization communication system according to claim 4, characterized in that: Add 30dB of Gaussian noise to the encrypted signal.
6. The LSTM-based laser chaotic synchronization communication system according to claim 4 or 5, characterized in that: The mixing ratio of the chaotic carrier signal and the information optical signal is set to 0.
08.
7. The LSTM-based laser chaotic synchronization communication system according to claim 4 or 5, characterized in that: The encrypted signal and the chaotic carrier signal are used as the input variable and the target output variable of the neural network in the LSTM network module (11), respectively. The neural network updates the state of its network nodes in each iteration until the ideal loss value is reached.
8. The LSTM-based laser chaotic synchronization communication system according to claim 7, characterized in that: The neural network structure in the LSTM network module (11) contains 200 LSTM nodes, a dropout layer with a dropout rate of 0.5, and a fully connected layer, wherein the training optimizer used is adam.