Abstract:Most existing DOA estimation methods assume ideal source incident angles with minimal noise. Moreover, directly using pre-estimated angles to calculate weighted coefficients can lead to performance loss. Thus, a green multi-modal (MM) fusion DOA framework is proposed to realize a more practical, low-cost and high time-efficiency DOA estimation for a H$^2$AD array. Firstly, two more efficient clustering methods, global maximum cos\_similarity clustering (GMaxCS) and global minimum distance clustering (GMinD), are presented to infer more precise true solutions from the candidate solution sets. Based on this, an iteration weighted fusion (IWF)-based method is introduced to iteratively update weighted fusion coefficients and the clustering center of the true solution classes by using the estimated values. Particularly, the coarse DOA calculated by fully digital (FD) subarray, serves as the initial cluster center. The above process yields two methods called MM-IWF-GMaxCS and MM-IWF-GMinD. To further provide a higher-accuracy DOA estimation, a fusion network (fusionNet) is proposed to aggregate the inferred two-part true angles and thus generates two effective approaches called MM-fusionNet-GMaxCS and MM-fusionNet-GMinD. The simulation outcomes show the proposed four approaches can achieve the ideal DOA performance and the CRLB. Meanwhile, proposed MM-fusionNet-GMaxCS and MM-fusionNet-GMinD exhibit superior DOA performance compared to MM-IWF-GMaxCS and MM-IWF-GMinD, especially in extremely-low SNR range.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the limitations of existing DOA (Direction of Arrival) estimation methods in practical applications. Specifically:
1. **Existing methods assume ideal conditions**: Most existing DOA estimation methods assume that the incident angle of the signal source is ideal and the noise is extremely small. However, in the actual communication environment, noise always exists, and this ideal assumption is difficult to achieve.
2. **Performance loss problem**: Directly using the estimated angle to calculate the weighting coefficient may lead to performance loss. For example, when using the angles estimated by advanced DOA methods (such as root - MUSIC or deep learning) to calculate the weighting coefficient, errors may be introduced, thus affecting the final DOA estimation accuracy.
3. **Performance improvement under low SNR and a small number of snapshots**: In the case of low signal - to - noise ratio (SNR) and a small number of snapshots, traditional DOA estimation methods often cannot provide high - precision estimation results.
To solve these problems, this paper proposes a Green Multi - modal (MM) fusion DOA framework, aiming to achieve more practical, low - cost, and high - time - efficiency DOA estimation. The specific contributions are as follows:
1. **Establish a multi - modal fusion DOA framework**: Based on the heterogeneous hybrid MIMO receiving array, a multi - modal fusion DOA framework is proposed to achieve low - delay, high - precision, and more realistic DOA measurement. This framework is implemented through three main steps:
- Use the Root - MUSIC algorithm to obtain the rough DOA values and the candidate angle sets from all sub - array groups of the FD sub - array and the H2AD array respectively.
- Based on the initial sample points provided by the FD sub - array, use the clustering method to distinguish the real angle classes in the candidate angle set.
- Fuse the two parts of the real angles to improve the DOA estimation performance.
2. **Propose an effective clustering method**: In order to obtain more accurate real solutions from the candidate angle set, a Global Maximum Similarity (GMaxCS) clustering method is proposed. This method significantly enhances the discrimination of direction information by directly reflecting the angles between the signal direction vectors. In addition, a Multi - modal Iterative Weighted Fusion (MM - IWF) method based on co - learning assistance is proposed to fuse the two parts of the real angles. The experimental results show that the proposed MM - IWF - GMaxCS method is superior to the MM - IWF - GMinD method.
3. **Introduce a fusion network (fusionNet)**: To further improve the accuracy of DOA estimation, a co - learning - assisted multi - modal fusion network (fusionNet) based on a three - layer fully - connected neural network (FCNN) is proposed. This network adaptively learns the fusion weights, which are independent of the ideal DOA angles, so it is more suitable for DOA estimation in practical scenarios. The simulation results show that the four proposed clustering - fusion methods (MM - IWF - GMinD, MM - IWF - GMaxCS, MM - fusionNet - GMinD, MM - fusionNet - GMaxCS) can approach the ideal DOA performance. Especially in the extremely low SNR range, MM - fusionNet - GMinD and MM - fusionNet - GMaxCS show better DOA estimation performance than other methods.
Through these improvements, this paper aims to develop a more practical, efficient, and robust passive DOA estimator that can provide high - precision DOA estimation in complex noise environments.