基于声学脉冲响应的室内建图算法研究

Research on indoor mapping algorithm based on acoustic impulse response

由于室内环境下噪声及混响干扰造成声学脉冲响应(Acoustic Impulse Response,AIR)的波达时间(Time of Arrival,TOA)模糊,导致现有室内空间几何建图技术存在计算复杂等缺点,提出一种基于AIR可信度判断与霍夫变换结合的定位和建图一步完成的方法。该方法基于到达时间差(Time Difference of Arrival,TDOA)迭代估计声源位置并计算准确的TOA,随后基于声收发设备与一阶反射TOA关于反射面的椭圆约束关系,结合可信度判断与霍夫变换,筛选可信低阶TOA以实现室内反射面的二维位置估计。蒙特卡洛仿真结果表明,在混响时间T60为0.1347 s、AIR误差标准差为22.7×10^-3 ms、大小为10 m×8 m×4 m复杂室内环境下,墙面定位平均距离误差为10.1 cm,平均角度误差为2.7958°。在大小为5.26 m×3.5 m×3.35 m的真实房间中,该方法可以实现声源与墙体位置的同时定位,完成规则盒型房间的二维地图重构,且墙面定位平均距离误差为2.6 cm,平均角度误差为2.17°。

The existing indoor mapping technology has the shortcomings in high cost and high computational complexity,which is caused by the TOA(time of arrival)ambiguity of the acoustic impulse response(AIR)due to noise and reverberation in the indoor environment.To solve this problem,an indoor localization and mapping method based on combining AIR credibility judgment and Hough transform is proposed in this paper.Firstly,the method iteratively estimates the position of the sound source based on the Time Difference of Arrival(TDOA)and calculates the accurate TOA.Then,by using the elliptical constraint relationship between the acoustic transceiver and the first-order reflection TOA on the reflective surface and combining the credibility judgment and Hough Transform,the reliable low-order TOA is selected to achieve the two-dimensional position estimation of the indoor reflective surface.Monte Carlo simulation results show that in the complex indoor environment of 10m×8m×4m with reverberation time T60 of 0.1347 s and AIR error standard deviation of 22.7×10^-3 ms,the average distance error of wall localization is 10.1cm,and the average angular error is 2.7958°.In the following experiments in a real room of 5.26 m×3.5 m×3.35 m,this method can realize the simultaneous localization of sound source and wall position,and complete the two-dimensional reconstruction of a box-type room.And the average distance error of wall localization is 2.6 cm,and the average angular error is 2.17°.