Research On Key Technologies Of Indoor Geometric Reconstruction With Microphone Array Based On Acoustic Impulse Response

At present,indoor positioning techniques based on various sensors have been matured to provide user location information for indoor location services.However,the development of these applications is based on the known indoor map.For the unknown indoor geometry,the indoor map can be obtained by the indoor geometric reconstruction technique of the microphone array based on Acoustic Impulse Responses?AIRs?.The time of Arrival?TOA?ambiguity of AIRs caused by the noise and reverberation interference in indoor environment leads to the cumbersome measurement and high computational complexity of the existing indoor geometric reconstruction technique of microphone array.In this paper,an indoor geometric reconstruction technique for estimating the reflective surface based on elliptical constraint model is proposed.The method iteratively locates the sound source based on the Time Difference of Arrival?TDOA?and calculates the accurate TOAs of each-order reflection.Based on the elliptical constraint relationship between the acoustic transceiver and the TOA of first-order reflection on the reflector,the trusted low-order TOAs are screened to realize the two-dimensional localization of the indoor reflecting surface,using credibility judgment and Hough transform algorithm.The research contents are summarized as follows:?1?For the problem that it is difficult to detect and extract multipath TOAs from AIRs in reverberation environment,an Exponentially-swept Sine Signal?ESS?with certain anti-interference ability is designed.The pulse waveform to be detected in the AIRs envelope is further highlighted by subtracting the fitting curve from the envelope.The multipath TOAs is extracted by the peak detection with threshold.?2?Considering the influence of asynchronous sound source and microphone array distribution on the accurate localization of sound source and reflector,the distributed microphone array is used to expand the spatial distribution range of the reflection points.The Levenberg-Marquardt sound source localization algorithm based on TDOA is used to estimate the position of the sound source iteratively.It can provide more accurate multipath reflection time delay estimation for subsequent algorithms.?3?The trusted first-order multipath TOAs from AIRs are screened based on the position relationship between the reflector and the multipath elliptical constraint and credibility judgment.The influence of noise and irrelevant TOAs information is further eliminated by combining Hough transform and least square method.The estimation of classical COmmon Tangent Algorithm?COTA?,which is susceptible to noise interference and can only estimate single reflector,is modified to provide more accurate reconstruction results.?4?The indoor geometric reconstruction technique is proposed in this paper.The technique and its reconstruction performance are verified in simulation of different signal environments and real room.Monte Carlo simulation results show that in the complex indoor environment of 10×8×4m with reverberation time RT₆₀ of 0.1347s and AIRs error standard deviation of 22.7×10^-3ms,the average distance error of wall localization is10.1cm,and the average angular error is 2.7958°.In the following experiments in a real room of 5.26×3.5×3.35m,this method can progressively realize the localization of sound source and wall,and complete the two-dimensional reconstruction of a box-type room.The average distance error of wall localization reduces from 4.38cm to 2.6 cm,and the average angular error reduces from 2.89°to 2.17°.