Loading...

Joint SourceChannel Coding in Video Transmission Using Deep Learning

Ghayoumi Ghamsari, Moein | 2021

1170 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 53718 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Behroozi, Hamid; Hossein Khalaj, Babak
  7. Abstract:
  8. In the signal transmission cycle in a telecommunication network, there are two coding blocks, which can be realized in two ways. In the traditional way, source coding is first used to remove redundancies and compress information. Then, channel coding is used to transmit on the telecommunication channel and deal with noise and other destructive factors of the channel. In other words, source coding and channel coding are done separately. In contrast, signal transmission can be done by joint source¬channel coding. According to Shannon separation theorem and with the fulfillment of the conditions mentioned in the theorem, the method of separation of source and channel coding in a point to point telecommunication channel achieves the optimal answer, but this method is not necessarily optimal in telecommunication networks. For this reason, several studies have been devoted to joint source¬channel coding.Over the years, many methods have been proposed by deep learning in various telecommunications sectors such as coding. One method that is fundamentally different from conventional coding methods is the use of autoencoder to model the communications system. This method was proposed in 2019 by Erina Bourtsoulatze and Denise Gunduz. In this method, joint source¬channel coding is used to send the image, and the autoencoder network is designed using convolutional layers. In this thesis, we generalize this method for sending video. As we know, the video signal consists of frames that adjacent frames are very similar to each other. Therefore, it is necessary to use this feature in network design. For this purpose, we use motion estimation and motion compensation structures. Another tool that we use to better process the video signal is the ConvLSTM layer, which we use in the structure of autoencoder, which is used to process video frames in groups. In this video transmission network, all components are designed with deep learning structures and the network design is independent of the channel. In this coding, instead of first compressing the source and then preparing the compressed output with the channel coding to be sent, the frame pixels are converted directly to the channel input values and are therefore simpler than conventional coding methods. A video dataset is prepared for the simulation, and the results are compared with the network results presented by Erina Bourtsoulatze and Denise Gunduz.Our proposed network has achieved better results in the PSNR criterion of 3dB and in the MS¬SSIM criterion of 0.03
  9. Keywords:
  10. Deep Networks ; Autoencoder ; Joint Source and Channel Coding ; Video Data ; Channel Coding ; Source Coding

 Digital Object List

 Bookmark

...see more