Research on key technologies of artificial-intelligence-based data denoising

With the rapid development of sensor networks, the Industrial Internet, and multimedia applications, data noise has become a major bottleneck limiting the accuracy of downstream intelligent analysis and decision-making. Traditional statistical filtering and signal processing methods struggle to handle high-dimensional, nonlinear, and non-stationary noise. In recent years, artificial intelligence (AI) technologies, leveraging deep representation, end-to-end learning, and self-supervised paradigms, have provided new approaches for data denoising in complex scenarios. This paper systematically reviews the latest research progress in image, audio, and time-series data denoising, with a focus on key technologies based on deep residual attention networks, self-supervised learning, Transformer architectures, and diffusion models. A unified framework for comparative experiments was established, a set of standardized evaluation frameworks and reproducible code libraries were contributed, and future research directions were discussed, providing valuable references and practical guidance for researchers and engineers in related fields.