研究目的
To improve the accuracy of Convolutional Neural Networks (CNN) based demodulator for Orbital Angular Momentum (OAM) demodulation using Conditional Generative Adversarial Networks (CGAN).
研究成果
The proposed OAM demodulation technique based on CGAN improves the accuracy of CNN-based demodulator from 93.56% to 98.36% under the specified conditions. This method provides an effective way to train an OAM demodulator.
研究不足
The study is limited by the turbulence intensity and the size of the dataset. The performance of the proposed method may vary under different turbulence conditions or with larger datasets.
1：Experimental Design and Method Selection:
The study proposes an OAM demodulation method based on CGAN to improve the accuracy of CNN-based demodulator. A CGAN is trained on a limited dataset, and the discriminator in CGAN is fine-tuned as a new classifier for OAM demodulation.
2：Sample Selection and Data Sources:
16 OAM modes are transmitted on the communication system under the turbulence intensity of 4×10-13m-2/3, and 1000 intensity maps of each mode are collected as a train dataset.
3：List of Experimental Equipment and Materials:
The study uses a spatial light modulator (SLM) to modulate continuous Gaussian beams into LG beams, which are transmitted to the receiver via the free space channel. The intensity images of the distorted LG beams are captured by a CCD camera.
4：Experimental Procedures and Operational Workflow:
The generator of CGAN is used to learn a distribution from training data, and the extra information C is used to control the class of OAM states. The discriminator is trained to distinguish between real data and fake data generated by the generator.
5：Data Analysis Methods:
The discriminator of CGAN is used as a feature extractor, and the number of neurons in the fully connected layers is modified to get the classifier used for OAM demodulation, called TestNet. The accuracy of the TestNet is evaluated on a test set of 1000 new images for each OAM mode.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
