General speech restoration demands techniques that can interpret complex speech structures under various distortions. While State-Space Models (SSMs) like SEMamba have advanced the state-of-the-art in speech denoising, they are not inherently optimized for critical speech characteristics like spectral periodicity or multi-resolution frequency analysis. In this work, we introduce an architecture tailored to incorporate speech-specific features as inductive biases. In particular, we propose Frequency GLP, a frequency feature extraction block that effectively and efficiently leverages the properties of frequency bins. Then, we design a multi-resolution parallel time-frequency dual processing block and a learnable mapping to further enhance model performance. With all our ideas combined, the proposed method SEMamba++ achieves the best performance among multiple baseline models on both seen and unseen datasets. Furthermore, the model is computationally efficient, making it a viable solution in resource-constrained settings.

Figure 1: Overall architecture of SEMamba++ (a) and Frequency GLP (b). In (a), "Mag.", "Comp.", and "Decomp." refer to magnitude, compression, and decompression, respectively. "Conv" and "TrConv" indicate the 2d convolution and transposed convolution that down- and upsample along frequency axis.