State Space and Self-Attention Collaborative Network with Feature Aggregation for DOA Estimation
Qi You, Qinghua Huang, Yi-Cheng Lin
TL;DR
This work tackles the challenge of accurate and efficient DOA estimation in dynamic, reverberant environments. It introduces FA-Stateformer, a hybrid architecture that fuses feature aggregation with a bidirectional state-space backbone (Bi-Mamba+) and a lightweight SEConformer to capture global and local temporal dependencies. Key contributions include a dedicated feature aggregation block, a compressed feed-forward SEConformer, time-shift convolution for extended receptive fields, and a Bi-Mamba+ module that collaborates with self-attention in a unified Stateformer. Empirical results on simulated data and LOCATA demonstrate superior accuracy and efficiency compared with strong baselines, with ablations validating the impact of each component. The approach offers a practical solution for real-time, multi-source DOA estimation in challenging acoustic environments and motivates extensions to broader audio tasks and domain-robustness strategies.
Abstract
Accurate direction-of-arrival (DOA) estimation for sound sources is challenging due to the continuous changes in acoustic characteristics across time and frequency. In such scenarios, accurate localization relies on the ability to aggregate relevant features and model temporal dependencies effectively. In time series modeling, achieving a balance between model performance and computational efficiency remains a significant challenge. To address this, we propose FA-Stateformer, a state space and self-attention collaborative network with feature aggregation. The proposed network first employs a feature aggregation module to enhance informative features across both temporal and spectral dimensions. This is followed by a lightweight Conformer architecture inspired by the squeeze-and-excitation mechanism, where the feedforward layers are compressed to reduce redundancy and parameter overhead. Additionally, a temporal shift mechanism is incorporated to expand the receptive field of convolutional layers while maintaining a compact kernel size. To further enhance sequence modeling capabilities, a bidirectional Mamba module is introduced, enabling efficient state-space-based representation of temporal dependencies in both forward and backward directions. The remaining self-attention layers are combined with the Mamba blocks, forming a collaborative modeling framework that achieves a balance between representation capacity and computational efficiency. Extensive experiments demonstrate that FA-Stateformer achieves superior performance and efficiency compared to conventional architectures.