OFDM-guided Deep Joint Source Channel Coding for Wireless Multipath Fading Channels

Mingyu Yang,Chenghong Bian,Hun-Seok Kim

DOI: https://doi.org/10.48550/arXiv.2109.05194

2021-09-11

Abstract:We investigate joint source channel coding (JSCC) for wireless image transmission over multipath fading channels. Inspired by recent works on deep learning based JSCC and model-based learning methods, we combine an autoencoder with orthogonal frequency division multiplexing (OFDM) to cope with multipath fading. The proposed encoder and decoder use convolutional neural networks (CNNs) and directly map the source images to complex-valued baseband samples for OFDM transmission. The multipath channel and OFDM are represented by non-trainable (deterministic) but differentiable layers so that the system can be trained end-to-end. Furthermore, our JSCC decoder further incorporates explicit channel estimation, equalization, and additional subnets to enhance the performance. The proposed method exhibits 2.5 -- 4 dB SNR gain for the equivalent image quality compared to conventional schemes that employ state-of-the-art but separate source and channel coding such as BPG and LDPC. The performance further improves when the system incorporates the channel state information (CSI) feedback. The proposed scheme is robust against OFDM signal clipping and parameter mismatch for the channel model used in training and evaluation.

Signal Processing

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to improve the transmission performance through joint source - channel coding (JSCC) when transmitting images in wireless multipath fading channels. Specifically, the paper proposes a deep - learning - based method, combined with orthogonal frequency - division multiplexing (OFDM) technology, to address the challenges brought by multipath fading and optimize the entire system in end - to - end training. The main objectives of the paper are:

1. **Improve image transmission quality**: By jointly optimizing source coding and channel coding, improve the transmission quality of images in multipath fading channels, especially under low signal - to - noise ratio (SNR) conditions.
2. **Reduce bit error rate**: By introducing explicit channel estimation and equalization methods, as well as additional sub - networks, further improve the performance of the system and reduce the bit error rate.
3. **Enhance robustness**: Make the system more robust to OFDM signal clipping, noise power variation and unseen channel conditions.
4. **Utilize channel state information (CSI) feedback**: Through precoding schemes, use channel state information for adaptive power allocation to further improve performance.

The paper achieves these goals through the following methods:

- **Combine OFDM and deep learning**: Use convolutional neural networks (CNN) as encoders and decoders to directly map the source image to complex - valued baseband samples for OFDM transmission.
- **Explicit channel estimation and equalization**: Introduce explicit channel estimation and equalization modules in the decoder and use domain knowledge to improve performance.
- **Residual connection**: Introduce two lightweight networks (CE sub - network and EQ sub - network) to compensate for the residual errors in channel estimation and equalization.
- **Adversarial loss**: Combine the adversarial loss of the generative adversarial network (GAN) to improve the perceptual quality of the image.

Through these methods, the paper shows that the proposed JSCC - OFDM method is significantly superior to the traditional separate source - channel coding schemes on a variety of test datasets.