Wireless Semantic Transmission via Revising Modules in Conventional Communications Peiwen Jiang Chao-Kai Wen Shi Jin and Geoffrey Ye Li

2025-04-29 0 0 1.37MB 16 页 10玖币

侵权投诉

Wireless Semantic Transmission via Revising

Modules in Conventional Communications

Peiwen Jiang, Chao-Kai Wen, Shi Jin, and Geoffrey Ye Li

Abstract

Semantic communication has become a popular research area due its high spectrum efﬁciency

and error-correction performance. Some studies use deep learning to extract semantic features, which

usually form end-to-end semantic communication systems and are hard to address the varying wireless

environments. Therefore, the novel semantic-based coding methods and performance metrics have been

investigated and the designed semantic systems consist of various modules as in the conventional

communications but with improved functions. This article discusses recent achievements in the state-of-

art semantic communications exploiting the conventional modules in wireless systems. We demonstrate

through two examples that the traditional hybrid automatic repeat request and modulation methods can

be redesigned for novel semantic coding and metrics to further improve the performance of wireless

semantic communications. At the end of this article, some open issues are identiﬁed.

I. INTRODUCTION

Semantic communication can signiﬁcantly reduce the requirements of transmission resources.

Different from conventional communications, semantic-based methods commonly rely on a

knowledge base (KB) to remove redundancy and correct errors during transmission. The KB can

be represented by a speciﬁc content or a set of trainable nerual networks. Compared to bit-level

transmission in the conventional communications, semantic communications are content-related

and directly transmits the desired meaning. Deep learning (DL)-based semantic communication

P. Jiang and S. Jin are with the National Mobile Communications Research Laboratory, Southeast University, Nanjing 210096,

China (e-mail: PeiwenJiang@seu.edu.cn; jinshi@seu.edu.cn).

C.-K. Wen is with the Institute of Communications Engineering, National Sun Yat-sen University, Kaohsiung 80424, Taiwan

(e-mail: chaokai.wen@mail.nsysu.edu.tw).

G. Y. Li is with the Department of Electrical and Electronic Engineering, Imperial Colledge London, London, UK (e-mail:

geoffrey.li@imperial.ac.uk).

arXiv:2210.00473v1 [eess.SP] 2 Oct 2022

usually realizes a content-related coding and decoding based on the same KB in an end-to-

end (E2E) manner. A proper KB is essential to the spectrum efﬁciency and error-correction

performance of a semantic system. Therefore, domain adaption [1] and federated learning [2]

methods have been developed to update and share a new KB for both the transmitter and the

receiver.

Recently, semantic communication is implemented by redesigning or revising the modules in

the conventional communications, which can be better address varying wireless environments.

In addition to semantic coding and decoding, other modules in the conventional communication

systems also need to be adjusted due to the change of transmission contents from symbol

sequence to semantic meaning and performance metrics. For semantic communications, the

modulation method is redesigned in [3] to maximize the sentence similarity rather than to

minimize the bit errors. This metric change signiﬁcantly affects the modulation because the words

with similar meaning can be modulated into close constellation points. In [4], peak-to-average-

power ratio (PAPR) is also reduced with semantic coding together to improve the semantic

similarity between the received and transmit sentences. Hybrid automatic repeat request (HARQ)

is a key technique to address varying wirelss channels in the conventional communications, which

has been exploited to develop the semantic-based HARQ in [5]. Furthermore, the semantic

transmitter can adaptively carry different amounts of semantic contents according to the channel

information [6]. To accommodate different performance metrics and requirements, the resource

allocation [7] for multi-user wireless communications becomes heterogeneous; therefore, the

complexity is sharply increased. In general, the novel performance metrics and transmission

methods for semantic communications require a brand-new design for wireless communications.

Different from the existing survey or tutorial literature, such as [8]–[10], this article focuses on

wireless semantic transmission based on revising or redesigning the modules in the conventional

communications. We ﬁrst look at the conventional modules as in a wireless communication

system in Fig. 1and then discuss its limitation. Then, the novel changes to facilitate semantic

transmission are described. In general, semantic communications thoroughly reform the trans-

mission paradigm as demonstrated by two examples in this article. Since the development of

semantic communications is still in its infant phase, we highlight some challenges on practical

wireless semantic communications.

The rest of this article is organized as follows. Section II introduces a conventional wireless

communication system and indicates the growth trend of wireless terminals and multimodal

requirements. Section III describes how semantic transmission affects the design of different

modules in communication systems. Section IV presents two examples on semantic channel

coding and modulation. Section V provides some open issues on wireless semantic communi-

cations. Section VI concludes this paper.

II. WHY WIRELESS SEMANTIC COMMUNICATION?

In this section, a conventional communication system is introduced ﬁrst. Then, the new

communication scenarios and requirements are discussed. Finally, the limitation of conventional

modules are pointed out.

Source

encoder

Source

decoder

Channel

encoder

Channel

decoder

Modulation Add

Pilot

Signal

detection Channel

estimation

IFFT&

add CP

Remove CP

&FFT

Channel

Segmentation

Knowledge

Base

Extraction

Generator

Joint channel

encoder

Joint channel

decoder

Modulation

face

ear

body

face

error

body

cat

tiger

dog

cat

tiger

dog

Signal

detection

Repaired

ear

“I eat a red apple”

User1

User2

11101... 11101...01... 1111

0010

11101... 10101...01...

Training data

...

Implicit

(Trained parameters)

Explicit

(Typical samples)

Semantic

extraction

Semantic

reconstruction

Semantic

Metrics

Semantic methods

guided by KB

Sentence similarity

BERT BERT

Transmit

sentence Received

sentence

Distance between

word embeddings

Pretrained

networks

...

Billions of

sentences

Pretrained model:

BERT

Extracting

semantic correlation

Semantic

Metric

Resource

allocation

Fig. 1. A conventional OFDM wireless communication systems. The green modules can be

potentially redesigned for semantic communications while the blue modules are still conventional.

A. A classic wireless communication system

We use orthogonal frequency division multiplexing (OFDM) as an example to discuss the

difference between the conventional and the semantic communication systems since OFDM is

widely used. As shown in Fig. 1, the source content, such as images or texts, is compressed

and converted into a bit sequence by a source encoder and then redundancy is added to the bit

sequence by a channel encoder to cope with the channel distortion. Next, a proper modulation

converts the bit sequence into a complex symbol sequence. The pilot is inserted for channel

estimation before inverse fast Fourier transform (IFFT). To facilitate OFDM demodulation,

cyclic preﬁx (CP) is added before sending to wireless channels. For multi-user networks, limited

wireless resources, such as bandwidth and transmission power, should be properly allocated to

optimizing network performance.

Many modules at the receiver, such as demodulation, channel decoding, and source decoding,

perform inverse operations of the corresponding modules at the transmitter. As shown in Fig. 1,

the FFT operation, channel estimation, and signal detection deal with the impact of channels.

The errors in the detected bits are corrected by channel decoder and the source decoder restores

the transmit content, such as image.

B. Growing demand of wireless services

Mobile work and online conferencing become essential parts of our life, especially during

the pandemic of COVID-19. For example, the transmission trafﬁc has been increased over 60%

compared to that before the outbreak of COVID-19. In order to deal with the unbearable demand,

some service providers, such as YouTube, can only reduce video qualities at peak times. On the

other hand, the users expect to enjoy a high-quality service, such as high-resolution videos,

without restrictions on time and place. As a result, semantic communication, which signiﬁcantly

improves transmission efﬁciency and enhance user’s experience, is desired.

Apart from improving the user’s experience, wireless networks also need to serve a huge

number of terminals. For example, autonomous cars rely on thousands of sensors for data

collection and cooperate with other vehicles. The data transmission in vehicular networks usually

serves speciﬁc tasks, where semantic communications are expected to play an important role.

C. Limitation of separate module design

In the classic Shannon’s paradigm, the channel coding has no need to consider the semantic

meaning of the transmit content. Thus, the conventional modules follow the divide-and-conquer

designs. However, the code length is limited in low-delay scenarios, such as conferencing and

autonomous drive. On the other hand, the transmission features under a speciﬁc task has strong

correlation. Thus, focusing on bit-level transmission is not efﬁcient any more in these situations.

The content-related semantic methods are brought to the forefront.

III. DEEP SEMANTIC SYSTEM DESIGNS

Most state-of-art works on semantic communication focus on the joint source-channel coding

(JSCC) design. Some methods redesign the modules in the conventional communication systems,

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

WirelessSemanticTransmissionviaRevisingModulesinConventionalCommunicationsPeiwenJiang,Chao-KaiWen,ShiJin,andGeoffreyYeLiAbstractSemanticcommunicationhasbecomeapopularresearchareadueitshighspectrumefciencyanderror-correctionperformance.Somestudiesusedeeplearningtoextractsemanticfeatures,whichusually...

展开>> 收起<<

Wireless Semantic Transmission via Revising Modules in Conventional Communications Peiwen Jiang Chao-Kai Wen Shi Jin and Geoffrey Ye Li.pdf

共16页,预览4页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

Wireless Semantic Transmission via Revising Modules in Conventional Communications Peiwen Jiang Chao-Kai Wen Shi Jin and Geoffrey Ye Li

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: