Back to Search
Start Over
Label-Assisted Transmission for Short Packet Communications: A Machine Learning Approach
- Source :
- IEEE Transactions on Vehicular Technology. 67:8846-8859
- Publication Year :
- 2018
- Publisher :
- Institute of Electrical and Electronics Engineers (IEEE), 2018.
-
Abstract
- Short packet communications (SPC) will play an important role in future Internet-of-Things networks. Conventional pilot-assisted transmission (PAT) needs significant overhead to obtain accurate channel state information (CSI) for further symbol detection and bit recovery, thereby reducing the spectral efficiency of the transmission. In this paper, a machine learning framework called Label-Assisted Transmission is proposed, in which the received signals are grouped into clusters through clustering algorithms and known labels are transmitted for cluster-symbol/bits mapping. This novel framework supports bit recovery directly without requiring the bit-symbol mapping information. When such mapping information is available, modulation constrained (MC) clustering algorithms are proposed, which exploit the unique characteristics of digital communication signals. For frequency-flat channels, this novel design needs only one known label regardless of the modulation size and the missing labels can be reconstructed using a proposed label reconstruction scheme. For frequency-selective channels with $L$ -tap time domain channel responses, only $L$ known labels are needed to reconstruct the missing labels if orthogonal frequency division multiplexing technology is adopted. The proposed clustering receiver works well even when the number of clusters is much larger than the number of received samples. The performance of the proposed framework is analyzed empirically through extensive simulations, which verify that the proposed scheme approaches the maximum likelihood detector with perfect CSI.
- Subjects :
- Computer Networks and Communications
Orthogonal frequency-division multiplexing
Computer science
Aerospace Engineering
02 engineering and technology
Machine learning
computer.software_genre
01 natural sciences
0203 mechanical engineering
Overhead (computing)
Electrical and Electronic Engineering
Cluster analysis
Computer Science::Information Theory
Network packet
business.industry
010401 analytical chemistry
020302 automobile design & engineering
Spectral efficiency
0104 chemical sciences
Transmission (telecommunications)
Modulation
Channel state information
Automotive Engineering
Artificial intelligence
business
computer
Communication channel
Subjects
Details
- ISSN :
- 19399359 and 00189545
- Volume :
- 67
- Database :
- OpenAIRE
- Journal :
- IEEE Transactions on Vehicular Technology
- Accession number :
- edsair.doi...........76e741849747ec6f05c3dcf3160230c8