Enhancing AI-Supported Channel Estimation in MIMO Systems with Open Set Recognition

Huiwen Zhang, Venkataramani Kumar, Feng Ye, Rose Qingyang Hu, Yi Qian

Abstract: Accurate channel estimation is required for various multiple input multiple output (MIMO) implementations in the next-generation wireless communication systems. Recently, Artificial Intelligence (AI) techniques have been introduced for channel state information (CSI) processing in channel estimation because of their accuracy and relatively low complexity compared to the traditional approaches. However, these AI-supported CSI processing models are usually developed with a fixed training dataset. Therefore, the performance of such approaches cannot be guaranteed in new environments. This paper focuses on enhancing AI-supported channel estimation methods with a 2-stage open set recognition scheme. New environments are detected in the first stage by identifying different characteristics between testing and training data. In the second stage, new data is filtered and further categorized to each individual environment. Simulation results using four different environment settings demonstrate that the proposed method can greatly enhance the usability of AI-supported channel estimation.

External IDs:dblp:conf/globecom/ZhangKYH024