Loading [MathJax]/extensions/MathMenu.js
Advancing Transparency in AI-based Automatic Modulation Classification | IEEE Conference Publication | IEEE Xplore

Advancing Transparency in AI-based Automatic Modulation Classification


Abstract:

Artificial Intelligence-based Automatic Modulation Classification has recently garnered significant attention due to advancements in AI technology. It has greatly improve...Show More

Abstract:

Artificial Intelligence-based Automatic Modulation Classification has recently garnered significant attention due to advancements in AI technology. It has greatly improved signal recognition in terms of accuracy and the range of detectable wireless signals. However, the inherent opacity of deep learning models presents challenges in understanding their internal workings and the basis for their predictions, making it difficult to ascertain whether the network has accurately learned the relevant features, thereby affecting the model’s reliability. In recent years, the topic of explainable deep learning technology has gained interest, with most methods primarily focusing on the more intuitive domain of image processing, leaving a notable gap in explainable artificial intelligence methods within the wireless signal domain. To address this gap, this paper explores the explainability of automatic modulation classification networks based on general CNN and LSTM architectures. By exploring various attribution methods, the aim is to visualize the key feature points learned by the network, which play a crucial role in classification predictions, making AMC’s outcomes more reliable and transparent.
Date of Conference: 07-09 August 2024
Date Added to IEEE Xplore: 24 September 2024
ISBN Information:
Print on Demand(PoD) ISSN: 2377-8644
Conference Location: Hangzhou, China

Funding Agency:


References

References is not available for this document.