Due to the rich lexical meanings and unique characteristics of the Chinese language, Chinese comments often lack explicit sentiment words, making implicit sentiment analysis challenging. This paper focuses on Chinese implicit sentiment analysis using the state-space model. We introduce a hierarchical feature network and forgetting gate residual structure to extract sentiment features at both the word and sentence levels, enabling sentiment classification. The main contributions are: (1) Adopting a state-space model architecture: We utilize the selective state-space model (Mamba) for efficient feature extraction, reducing training costs. (2) Introducing a hierarchical feature network: The HGRU network extracts word- and sentence-level features to identify sentiment. (3) Introducing an oblivious gate residual network: This improves model robustness. Experiments on three Chinese sentiment datasets (Weibo, chnsenticorp-htl, and, SMP-ECISA2019) show significant accuracy improvement.