Deep Joint Source-channel Coding for Wireless Image Transmission

By Eirina Bourtsoulatze, David Burth Kurka, Deniz Gündüz. IEEE International Conference on Acoustics, Speech, and Signal Processing, 2018

1,421 citations recorded by the source corpus.

We propose a novel joint source and channel coding (JSCC) scheme for wireless image transmission that departs from the conventional use of explicit source and channel codes for compression and error correction, and directly maps the image pixel values to the complex-valued channel input signal. Our encoder-decoder pair form an autoencoder with a non-trainable layer in the middle, which represents the noisy communication channel. Our results show that the proposed deep JSCC scheme outperforms separation-based digital transmission at low signal-to-noise ratio (SNR) and low channel bandwidth regimes in the presence of additive white Gaussian noise (AWGN). More strikingly, deep JSCC does not suffer from the "cliff effect" as the channel SNR varies with respect to the SNR value assumed during training. In the case of a slow Rayleigh fading channel, deep JSCC can learn to communicate without explicit pilot signals or channel estimation, and significantly outperforms separation-based digital communication at all SNR and channel bandwidth values.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Signal Processing

View Paper: https://www.semanticscholar.org/paper/40e4a21b6a4d55cacc57b96ee0d7674ed35f16ec

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Signal Processing

Citations in the corpus

A Lite Distributed Semantic Communication System for Internet of Things (2020, methodology): “[18] proposed joint source-channel coding for wireless image transmission based on the convolutional neural network (CNN), where peak signal-to-noise ratio (PSNR) is used to measure the accuracy of image recovery at the receiver.”
Machine Learning in the Air (2019, methodology): “Performance of the deep JSCC algorithm in [59] on CIFAR-10 test images transmitted over an AWGN channel with respect to the available channel bandwidth per image pixel for different SNR values.”
Beyond Transmitting Bits: Context, Semantics, and Task-Oriented Communications (2022, positive): “Recently, JSCC schemes based on autoencoders [200], which are DNNs aimed at unsupervised data coding, have been introduced [201]–[205] and are shown to provide comparable or better performance than state-of-the-art separation-based digital schemes.”
Wireless Image Transmission Using Deep Source Channel Coding With Attention Modules (2020, methodology): “Akin to [29]–[31], we call the image size n, the channel input size k and R = k/n as the source bandwidth, the channel bandwidth and bandwidth ratio, respectively.”
Semantic Communications: Principles and Challenges (2021, background): “Joint source-channel coding (JSCC) schemes [10], [11] were proposed to capture and transmit semantic features, in which the semantic receiver executes the corresponding actions directly rather than recovering the source messages.”
Nonlinear Transform Source-Channel Coding for Semantic Communications (2021, methodology): “…for training the proposed NTSCC model for large images consists of 500,000 images sampled from the Open Images Dataset [38] 2) Comparison Schemes: Our comparison schemes include both the emerging deep JSCC scheme proposed in [9] and the classical separation-based source and channel coding schemes.”
Deep Learning Enabled Semantic Communication Systems (2020, neutral): “Initially, natural language can be described by the joint probability model according to the context [24].”
A Very Brief Introduction to Machine Learning With Applications to Communication Systems (2018, background): “Examples of algorithm deficit include channels with complex non-linear dynamical models, such as optical links [65]; Gaussian channels with feedback, for which optimal practical encoding schemes are not known [66]; multiple access channels with sparse transmission codes [67]; and joint source-channe”
Resource Allocation for Text Semantic Communications (2022, neutral): “Particularly, different types of semantic systems have been studied for different types of sources, including text [3], [4], image [5], [6], speech [7], and video [8], to ensure signiﬁcant improvement in semantic transmission reliability.”
Learning Task-Oriented Communication for Edge Inference: An Information Bottleneck Approach (2021, methodology): “Following the experimental settings in Section V, we also adopt DeepJSCC as the baseline in comparison.”