Study On Wideband Sound Source DOA Methods Of Small Aperture Array

Wideband sound source localization technology,one of the main technologies in the field of array signal processing,offers a wide range of potential applications in the fields of human-computer interaction,traffic noise localization,and unmanned aerial vehicle identification.At low signal-to-noise ratios,the traditional wideband signal DOA(Direction of Arrival)algorithms suffer from low resolution and high time complexity.Furthermore,constrained by the size and structure of the mounted platform,the array aperture is difficult to be arbitrarily expanded for the array structure,so that the accuracy and resolution of wideband sound source DOA estimation algorithm are affected.Based on acoustic theory and array signal processing theory,studies on small-aperture arrays and wideband source DOA algorithms with low signal-to-noise ratio are conducted from theoretical analysis,numerical simulation and experimental testing to provide some theoretical and experimental references for the engineering implementation of wideband source DOA technology.Firstly,to address the trade-off between array structure and DOA performance,the impact of array structure on directional accuracy and array ambiguity using Cramer-Rao criteria and array ambiguity theory are explored.And a miniaturized three-dimensional array is proposed.Simulation experiments indicate that the DOA estimation accuracy based on such proposed array is higher and there is no Rank-1 ambiguity.Secondly,to address the issue that the test of orthogonality of subspace algorithms have pseudo-peaks and low resolution at low signal-to-noise ratios,an improved algorithm based on noise subspace reconstruction is proposed.The noise subspace is replaced with the covariance matrix to avoid eigenvalue decomposition.The computational complexity of the algorithm is reduced by calculating the spatial spectrum using the Frobenius norm of matrix rather than minimum singular value.The simulations under the uniform line array model and the proposed small-aperture stereo array model show that the proposed improved method could completely suppress the pseudo-peaks with the highest resolution at a low signal-to-noise ratio.Then,to overcome the problems of traditional coherent signal subspace algorithms,which require pre-estimation and are computationally intensive,a coherent wideband DOA algorithm based on energy weighting is presented.For the proposed algorithm,focusing matrix without pre-estimation is constructed and the optimal subbans are selected via energy to reduce the computational cost.Also,the decoherence capability is improved through convariance matrix reconstruction.Simulations on uniform line array show that the proposed algorithm has smaller estimation error and lower computational complexity at low signal-to noise ratio in comparision with the original algorithms.To solve the problem of high computational requirements in 2D search on the stereo array,a dimensionality reduction method is proposed.Through simulation,it is verified that such processes could enhance computational efficiency.Finally,experimental validation of the algorithm performance is carried out based on established experimental platforms,including a uniform line array and a proposed stereo array.Chirp signals,multi-frequency narrowband mixed signals,and voice signals are tested as sound sources for off-line.DOA estimations are determined offline using the proposed improved algorithms and conventional algorithms,respectively.The experimental results validate the usability of the experimental platform and the effectiveness of the proposed algorithms.