High Performance DOA Estimation Of Complex Indoor Environment Based On Acoustic Array

In the field of target monitoring in the context of the Internet of things(IoT),acoustic source direction finding,positioning and tracking using acoustic array technology has been an important research topic.A single acoustic array,which is composed of a plurality of microphones with high direction-finding accuracy and good concealment characteristics,is widely used in low al-titude UAV/helicopter warning,wild animal habitat monitoring,underwater torpedo/submarine positioning,tracking and other defense and civilian fields,and plays an important role.In recent years,with the demand for location of sound sources in smart home and indoor localization,indoor acoustic array technology has become a research focus in academia and industry.However,due to the limitation of the serious interference of indoor reverberation and noise,and the ambiguity of DOA estimation,how to achieve high-accuracy DOA estimation in complex indoor scenarios is a challenging task.This thesis is concentrated on indoor localization of sound sources using acoustic array technology,including studying an effective method of unwrapping ambiguity,finding a way for suppressing the reverberation and noise in DOA estimation,and implementing a single acoustic array localization system.The main contents are as follows:Firstly,the theoretical basis of acoustic array is studied,including the narrowband and wide-band array signal models,the evaluation criterion of array performance,and the different kinds of DOA estimation algorithms are carefully analyzed and derived.Besides,the basic theory of compressed sensing is studied,including sparse signal representation,measurement matrix design and signal reconstruction,and the DOA estimation algorithm based on compressed sensing(CS)sparse reconstruction is theoretically analyzed.Secondly,the theory of multi-band unwrapping ambiguity is proposed,and the constraint condition between frequency band and array interval is also deduced,which brings larger freedom to the choice of array interval,thus it can effectively improve the resolution of DOA estimation.DOA estimation algorithm based on CS is used for simulation verification.When the interval be-tween elements is larger than half wavelength,the true angle of arrival can be estimated as long as the ambiguity constraint is satisfied.At the same time,in order to realize the stable unwrapping ambiguity,the number of required frequency bands for unwrapping ambiguity under different S-NR is studied.In addition,we compare the unwrapping ambiguity algorithm based on CS DOA estimation with that based on MUSIC algorithm and Beamforming algorithm.It is found that the CS based DOA estimation algorithm has obvious advantages in angle resolution,avoiding the repetition of the spatial spectrum aliasing problem.Then,a time delay estimation algorithm based on frequency domain(FD-TDE algorithm)is proposed.The algorithm can effectively improve the performance of the estimation of the angle of arrival in the indoor environment.By establishing a signal error estimation model based on fre-quency domain,the expression of the FD-TDE algorithm is derived,and the unwrapping ambiguity characteristic of the method is analyzed based on the unwrapping ambiguity constraint.Based on the multi measurement information,the least square method is adopted to improve DOA estima-tion accuracy.The IMAGE Model is used to verify the performance of the FD-TDE algorithm inthe indoor environment of different SNR and reverberation(RT60).Simulation and experiments show that compared with GCC-PHAT algorithm,FD-TDE algorithm has better ability to suppress reverberation and noise,and it also has lower complexity and can achieve stable and accurate DOA estimation.Finally,the indoor localization system based on a single acoustic array and a smart phone are designed and implemented.Simulation results show that chirp signal has stronger ability of sup-pressing reverberation and noise than speech signal.At the same time,it is proposed that chirp can be dereverberation based on fractional Fourier domain,which further improves the performance of DOA estimation.In the experiment,the way of TPSN achieves distance measurement,which is similar to Timing-sync Protocol for Sensor Networks,and angle estimation is based on FD-TDE algorithm.In the indoor scenario(office,corridor),we compare the performance of high fre-quency chirp and low frequency chirp signal,analyzing the error of DOA estimation and indoor localization.