Research Of Digital Signal Processing Techniques Using Support Vector Machine For Free-Space Optical Communications

With the advancement of the information society, the demands for high-speed and large-capacity communications are also on the increase. And the bandwidth limitations of satellite microwave communication and "the last kilometer" problem of optical fiber communication are still great challenges for us. Thus, free space optical communication (FSO) has began to be taken seriously. It has become important technology to obtain and transmit the high-speed and large-capacity spatial information. Unfortunately, the atmospheric turbulence on the atmospheric channel can reduce the transmission reliability of FSO system and the communication quality is easily restricted. Undoubtedly, the suppression techniques of atmospheric turbulence are worth studying and have promising applications. At present, the digital signal processing (DSP) technology are also in the rapid development, and new DSP techniques have been applied in the field of FSO communication and created new research directions.In this thesis, we employ the support vector machine (SVM) algorithm as the DSP technique for FSO system to suppress the atmospheric turbulence effect and restore the original transmission signal. The main research works are as follows:1) Based on the FSO system model, we briefly introduce the basic theory of atmospheric turbulence and the transmission mechanism of the laser beam that propagating at the atmospheric channel, to lay the foundation for the numerical simulations of atmospheric turbulence and FSO system.2) The method of distributed Fourier phase screens is applied to simulate the atmospheric turbulence. Based on the spectrum inversion method, we generate the random phase screens with atmospheric turbulence characteristics by simulation software. Using the distributed Fourier phase screen technique, we place multiple random phase screens in the transmission path of laser beam, to simulate the transmission process of laser beam propagating at the atmospheric turbulence under certain conditions. To make up for the loss on the low-frequency component of phase screens, we also use the harmonics compensation method to compensate the low-frequency loss and further improve the simulation accuracy.3) Considering the fading at the turbulence channel is of millisecond level, we propose a large-block detection method based on the SVM algorithm. The block size of the proposed detection is of hundreds of thousands of bits or even more, and is proportional to the bit rate of FSO system. Compared with traditional detection method, the SVM scheme is not only independent of the channel’s turbulence condition and the fading model of receiving signal, but also requires no knowledge of prior information or heuristic assumptions. Numerical simulations show that the SVM algorithm is an effective DSP method to mitigate the atmospheric turbulence fading. When the receiver aperture is sufficiently large, the SVM detection can approach closely to the detection with prefect channel state information (CSI). In the last, we further consider the technique of space diversity and propose a combination scheme of SVM and MIMO to mitigate the turbulence-fading. The related numerical simulation shows that this scheme can achieve good performance.