DNN-based Overhead Reduction for High-Quality Soft Delivery

    •  Fujihashi, T., Koike-Akino, T., Watanabe, T., Orlik, P.V., "DNN-based Overhead Reduction for High-Quality Soft Delivery", IEEE Global Communications Conference (GLOBECOM), DOI: 10.1109/​GLOBECOM38437.2019.9014124, December 2019.
      BibTeX TR2019-133 PDF
      • @inproceedings{Fujihashi2019dec2,
      • author = {Fujihashi, Takuya and Koike-Akino, Toshiaki and Watanabe, Takashi and Orlik, Philip V.},
      • title = {DNN-based Overhead Reduction for High-Quality Soft Delivery},
      • booktitle = {IEEE Global Communications Conference (GLOBECOM)},
      • year = 2019,
      • month = dec,
      • publisher = {IEEE},
      • doi = {10.1109/GLOBECOM38437.2019.9014124},
      • issn = {2576-6813},
      • isbn = {978-1-7281-0962-6},
      • url = {}
      • }
  • MERL Contacts:
  • Research Areas:

    Communications, Digital Video, Signal Processing


Soft delivery, i.e., analog transmission, has been proposed to provide graceful video/image quality even in unstable wireless channels. However, existing analog schemes require a significant amount of metadata for power allocation and decoding operations. It causes large overheads and quality degradation due to rate and power losses. Although the amount of overheads can be reduced by introducing Gaussian Markov random field (GMRF) model, the model mismatch can degrade reconstruction quality. In this paper, we propose a novel analog transmission scheme to simultaneously reduce the overheads and yield better reconstruction quality. The proposed scheme uses a deep neural network (DNN) for metadata compression and decompression. Specifically, the metadata is compressed into few variables using the proposed DNN-based metadata encoder before transmission. The variables are then transmitted and decompressed at the receiver for high-quality video/image reconstruction. Evaluations using test images demonstrate that our proposed scheme reduces overheads by 80.0 % with 11.2 dB improvement of reconstruction quality compared to the existing analog transmission schemes.


  • Related News & Events

    •  NEWS    MERL Scientists Presenting 11 Papers at IEEE Global Communications Conference (GLOBECOM) 2019
      Date: December 9, 2019 - December 13, 2019
      Where: Waikoloa, Hawaii, USA
      MERL Contacts: Jianlin Guo; Kyeong Jin (K.J.) Kim; Toshiaki Koike-Akino; Philip V. Orlik; Pu (Perry) Wang
      Research Areas: Communications, Computer Vision, Machine Learning, Signal Processing, Information Security
      • MERL Signal Processing scientists and collaborators will be presenting 11 papers at the IEEE Global Communications Conference (GLOBECOM) 2019, which is being held in Waikoloa, Hawaii from December 9-13, 2019. Topics to be presented include recent advances in power amplifier, MIMO algorithms, WiFi sensing, video casting, visible light communications, user authentication, vehicular communications, secrecy, and relay systems, including sophisticated machine learning applications. A number of these papers are a result of successful collaboration between MERL and world-leading Universities including: Osaka University, University of New South Wales, Oxford University, Princeton University, South China University of Technology, Massachusetts Institute of Technology and Aalborg University.

        GLOBECOM is one of the IEEE Communications Society’s two flagship conferences dedicated to driving innovation in nearly every aspect of communications. Each year, more than 3000 scientific researchers and their management submit proposals for program sessions to be held at the annual conference. Themed “Revolutionizing Communications,” GLOBECOM2019 will feature a comprehensive high-quality technical program including 13 symposia and a variety of tutorials and workshops to share visions and ideas, obtain updates on latest technologies and expand professional and social networking.