| Presentation | 2020-03-04 CSI Feedback Overhead Reduction by 3D CNN for Time-varying FDD Massive MIMO Masumi Kuriyama, Tomoaki Ohtsuki, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Massive MIMO (Multiple-Input Multiple-Output) is a technology that uses a large number of antennas at a base station (BS), thereby achieving high communication performance. However, massive MIMO has a problem that the feedback of channel state information (CSI) required for precoding in the BS increases due to the large number of antennas. Recently, there is a technique of using deep learning to address this problem. In the conventional method using deep learning, features are extracted by treating CSI matrices represented by space and frequency as images, and are used for compression and reconstruction. In this report, we propose a method that uses a 3-dimensional convolutional neural network (3D CNN) to extract features in the time domain of CSI, in addition to the spatial and frequency domains, and to perform compression and reconstruction. Furthermore, the proposed method uses prediction by Convolutional LSTM (ConvLSTM), a kind of recurrent neural network (RNN), to compensate for the difference between the reconstructed CSI and that required for precoding due to the feedback delay of the time-varying channel. The proposed method achieves high accuracy by CSI compression /reconstruction using 3D CNN and channel prediction using ConvLSTM, even if information is compressed 1/64. Also, the proposed method improves the reconstruction accuracy at an arbitrary compression ratio compared to the conventional one that learns only two dimensions of space and frequency. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Massive MIMO / channel feedback / 3D CNN / ConvLSTM |
| Paper # | RCS2019-332 |
| Date of Issue | 2020-02-26 (RCS) |
| Conference Information | |
| Committee | RCS / SR / SRW |
|---|---|
| Conference Date | 2020/3/4(3days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | Tokyo Institute of Technology |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | Mobile Communication Workshop |
| Chair | Tomoaki Otsuki(Keio Univ.) / Masayuki Ariyoshi(NEC) / Satoshi Denno(Okayama Univ.) |
| Vice Chair | Satoshi Suyama(NTT DoCoMo) / Fumiaki Maehara(Waseda Univ.) / Toshihiko Nishimura(Hokkaido Univ.) / Suguru Kameda(Tohoku Univ.) / Osamu Takyu(Shinshu Univ.) / Kentaro Ishidu(NICT) / Keiichi Mizutani(Kyoto Univ.) |
| Secretary | Satoshi Suyama(NTT) / Fumiaki Maehara(Kyushu Univ.) / Toshihiko Nishimura(ATR) / Suguru Kameda(Univ. of Electro-Comm.) / Osamu Takyu(Mie Univ.) / Kentaro Ishidu(Tokyo Inst. of Tech.) / Keiichi Mizutani(Anritsu) |
| Assistant | Kazushi Muraoka(NEC) / Shinsuke Ibi(Doshisha Univ.) / Koichi Adachi(Univ. of Electro-Comm.) / Osamu Nakamura(Sharp) / Manabu Sakai(Mitsubishi Electric) / Mai Ohta(Fukuoka Univ.) / Teppei Oyama(Fujitsu Lab.) / Kentaro Kobayashi(Nagoya Univ.) / Masaaki Fuse(Anritsu) / Tomoki Murakami(NTT) |
| Paper Information | |
| Registration To | Technical Committee on Radio Communication Systems / Technical Committee on Smart Radio / Technical Committee on Short Range Wireless Communications |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | CSI Feedback Overhead Reduction by 3D CNN for Time-varying FDD Massive MIMO |
| Sub Title (in English) | |
| Keyword(1) | Massive MIMO |
| Keyword(2) | channel feedback |
| Keyword(3) | 3D CNN |
| Keyword(4) | ConvLSTM |
| 1st Author's Name | Masumi Kuriyama |
| 1st Author's Affiliation | Keio University(Keio Univ.) |
| 2nd Author's Name | Tomoaki Ohtsuki |
| 2nd Author's Affiliation | Keio University(Keio Univ.) |
| Date | 2020-03-04 |
| Paper # | RCS2019-332 |
| Volume (vol) | vol.119 |
| Number (no) | RCS-448 |
| Page | pp.pp.63-68(RCS), |
| #Pages | 6 |
| Date of Issue | 2020-02-26 (RCS) |