SNN-SC: A Spiking Semantic Communication Framework for Classification
Preprint
Abstract
Nowadays, deep learning (DL)-based semantic communication (SC) is gaining more and more attention. In SC, only the information that is important, relevant, and useful for downstream tasks is extracted from the source and delivered to the receiver. Most existing SC works are only designed, trained, and tested on Additive White Gaussian Noise (AWGN) channels to transmit analog signals, and only a few SC works are applied to digital channels and require additional quantization. In actual communication systems, digital signals are considered more. Therefore, it is necessary to design an end-to-end SC framework for digital communication systems. In this paper, we propose a spiking neural network (SNN) based digital SC framework, namely SNN-SC, to address this problem. SNN-SC is used to compress and extract the semantic information of classification model features in collaborative intelligence (CI) scenario, and the extracted semantic information is transmitted through digital channels such as binary symmetric channel (BSC) and binary erasure channel (BEC). Since the outputs of SNN are binary spikes, the framework can be directly applied to digital channels without extra quantization and channel coding. Furthermore, we propose a novel spiking neuron and regularization method to improve the performance and robustness of the framework. We evaluate SNN-SC on different classification models, and experimental results show that the proposed SNN-SC framework outperforms the state-of-the-art convolutional neural network (CNN)-based SC framework on different bandwidths and digital channels. In addition, the experimental results also demonstrate that SNN-SC exhibits good generalization ability on different classification models.Keywords
All Related Versions
This publication has 16 references indexed in Scilit:
- Deep Source-Channel Coding for Sentence Semantic Transmission With HARQIEEE Transactions on Communications, 2022
- Semantic Communications: Overview, Open Issues, and Future Research DirectionsIEEE Wireless Communications, 2022
- An adaptive threshold mechanism for accurate and efficient deep spiking convolutional neural networksNeurocomputing, 2022
- Deep Joint Source-Channel Coding for Multi-Task NetworkIEEE Signal Processing Letters, 2021
- Spiking Neural Networks—Part I: Detecting Spatial PatternsIEEE Communications Letters, 2021
- A Lite Distributed Semantic Communication System for Internet of ThingsIEEE Journal on Selected Areas in Communications, 2020
- Deep Joint Source-Channel Coding for Wireless Image TransmissionIEEE Transactions on Cognitive Communications and Networking, 2019
- Deep neural networks with weighted spikesNeurocomputing, 2018
- Spiking Deep Convolutional Neural Networks for Energy-Efficient Object RecognitionInternational Journal of Computer Vision, 2014
- Networks of spiking neurons: The third generation of neural network modelsNeural Networks, 1997