The Study On The Underdetermined DOA Estimation Based On Acoustic Vector Sensor

Super resolution estimation of direction of arrival(DOA) has been a hot issue in array signal processing research. In particular, the underdetermined DOA estimation has a definite application background and a wide range of actual demand, so it has become the hotspot and difficulty in research in recent years. In the field of DOA estimation, the number of targets resolved by the array is limited by the array elements, and the estimation performance of the algorithm can be seriously reduced when the number of targets exceeds the number of array elements. If the target numbers are far greater than the number of array elements, the high resolution DOA estimation can be realized, which is the ultimate goal of the underdetermined DOA estimation.An acoustic vector sensor is composed by three mutually vertical velocity sensors and a pressure sensor, and the collocated acoustic vector sensor can simultaneously measure the pressure and particle velocity of the same point in acoustic field. Unfortunately, this structure not only reduces the aperture of the array and the resolution of the sensor, but also increases the complexity of hardware, whose only benefit is to reduce the difficult of signal processing. But from the point of view of engineering, we are more willing to increase the difficulty of signal processing in exchange for lower hardware costs and higher estimation accuracy, and the spatial non-collocated acoustic vector sensor just meets our needs.In recent years, the one-dimensional DOA estimation algorithms are emerging, but in essence, the core idea of these algorithms constructs the virtual array, expands the array aperture and improves the degree of freedom. Due to the harsh conditions and the less methods of the one-dimensional DOA estimation algorithms and the high complexity of the two-dimensional DOA estimation algorithms, many one-dimensional DOA estimation algorithms are difficult to be directly applied to the two-dimensional DOA estimation algorithms. In addition, there is little algorithm model is directly aimed at the two-dimensional DOA estimation at present, so the two-dimensional DOA estimation is still an urgent need to be solved in the academic field. To overcome this difficulty, I decide to focus on the two-dimensional underdetermined DOA estimation. Through the study of the current underdetermined DOA estimation algorithm, an improved algorithm is proposed which is the underdetermined DOA estimation based on tensor model using acoustic vector sensor arrays. In the same condition, it can achieve higher estimation accuracy than the original algorithm and the advantage is obvious in the case of low SNR, but will not significantly increase the complexity of the algorithm. However, due to the limitations of the algorithm, the algorithm can only achieve the one-dimensional underdetermined DOA estimation and the two-dimensional overdetermined DOA estimation. In order to achieve the meaningful two-dimensional underdetermined DOA estimation, we propose a new L array DOA estimation algorithm based on tensor model and this algorithm can achieve the meaningful two-dimensional underdetermined high accuracy DOA estimation in the case of low SNR.The spatial non-collocated acoustic vector sensor can effectively reduce the hardware cost and improve the sensor estimation accuracy and so on. The two-dimensional underdetermined DOA estimation based on the spatial non-collocated acoustic vector sensor is still blank in the academic field. The realization of the two-dimensional underdetermined DOA estimation using the structural advantages of the spatial non-collocated acoustic vector sensor has important academic significance and practical value. The last part mainly explores two-dimensional underdetermined DOA estimation based on the spatial non-collocated acoustic vector sensor, and has solved a key problem. Although it has a series of advantages for the spatial non-collocated acoustic vector sensor, it introduces the problem of the "fuzzy". Some scholars present an effectively algorithm removing the "fuzzy", and the algorithm can achieve high accuracy DOA estimation. Because of the limitation of the algorithm itself, the performance of the proposed algorithm is drastically reduced due to the coherent sources. In addition, the coherent sources are introduced in the process of using a series of operations to expand the virtual aperture for the underdetermined DOA estimation, and will lead to the failure of the underdetermined DOA estimation. According to the above analysis, removing the coherence of the source is the key to achieve the two-dimensional underdetermined DOA estimation. In order to solve this problem, an improved algorithm is proposed. What is more, the proposed algorithm can obtain the equivalent accuracy of the original algorithm in the condition of the non-coherent sources. Under the condition of coherent sources, the original algorithm is almost failed, but the improved algorithm can still maintain a relatively high accuracy. The improved algorithm will lay a solid foundation for the realization of the two-dimensional underdetermined DOA estimation based on the spatial non-collocated acoustic vector sensor.