Research On Indoor Spatial Contour Building Based On Multi-Channel Pseudo-ultrasound Signals From Mobile Devices

The rapid development of location based service(LBS)related technologies has greatly promoted the construction process of smart cities.As the internal space of large building complexes such as airports,high-speed railway stations and shopping centers becomes increasingly large and complex,applications such as indoor positioning and navigation,path planning and virtual reality have created an urgent demand for indoor spatial map and location services.In response to the huge demand for indoor location and mapping services,Simultaneous Localization and Mapping(SLAM)technology has emerged,among which SLAM technology based on acoustic signals has received wide attention due to its low price,high ranging accuracy and good device compatibility.In this paper,we focus on the problem of first-order echo labeling in indoor reverberant conditions and the problem of observation error and cumulative error affecting the accuracy of map construction in the mobile measurement process.Combining acoustic spatial features and reflection echo identification and extraction technology,indoor ranging and positioning technology and multi-sensor fusion technology,we explore a low-cost solution for the rapid construction of spatial contour maps in indoor privacy environment.The main research of this paper is as follows:(1)Research on spatial sensing method based on multi-channel pseudo-ultrasound signal.Multi-channel microphones have better spatial perception capability than singlechannel microphones,and can discriminate the magnitude and direction of sound and even perceive the spatial contour of the room.The study of the spatial perception method of multi-channel pseudo-ultrasonic signals will be of great significance for the subsequent construction of the spatial contour of the room.The multi-channel microphone estimation of wall position needs to mark the first-order echoes from the reflected wall from the echo sequence received by the microphone to determine its own position relative to the reflector,however,the reflection source and arrival order of the first-order echoes received by the microphone are not consistent,which leads to an Echo Labelling Problem(ELP).This study addresses this problem by proposing a RIR cepstrum feature,which can be extracted from the Room Impulse Responses(RIR)of the acoustic signals received by the microphone array,and can not only identify the low-order echo signals,but also distinguish the reflectors of different materials in space.Based on this perceptual recognition property of RIR cepstrum,a first-order echo labeling algorithm based on RIR cepstrum is proposed to realize the perception of spatial sound field by multi-channel microphone and improve the recognition accuracy of first-order echo by establishing the association between pseudo-ultrasound data space and indoor spatial geometric features.(2)Research on pseudo-ultrasonic indoor geometric spatial contour reconstruction technique.During the propagation of pseudo-ultrasonic sound signal in the room,it will generate multipath effects such as reflection and diffraction when encountering obstacles,and such reflection and multipath propagation reflect the spatial layout and geometric structure information of the room to a certain extent.The mobile terminal can realize the reconstruction of indoor spatial contour by collecting the acoustic echo signals at different locations of the room and estimating the position of the reflecting walls.During the mobile terminal mobile measurement,the use of inertial sensors can provide better instantaneous positioning accuracy,but the accumulated errors in long-term mobile measurement can affect the accurate estimation of wall positions.The acoustic signal-based position estimation does not have error accumulation and high ranging accuracy,but due to the reverberation of acoustic signal during indoor propagation,it will lead to partial signal arrival detection omission or false detection.To address the above problems,this study integrates the complementary properties of inertial sensors and acoustic sensors and proposes a graph optimization-based acoustic SLAM method.Firstly,an isosceles trapezoidal constraint between the real acoustic source and the first-order mirror acoustic source position in the mobile terminal mobile measurement process is utilized to propose an isosceles trapezoidal constraint-based locus pose estimation method to estimate the mobile terminal’s mobile trajectory,which reduces the localization error of the indoor geometric contour by reducing the real acoustic source localization error.Then,the poses of the robot are treated as nodes of the graph,and the relationships among the poses constitute the edges of the graph,and the poses of the robot are corrected by global optimization using the constraint feedback of loopback detection.Finally,the geometric contour map of the room is drawn based on the globally optimized robot trajectory and the reflector discrimination property of the RIR inversion spectrum.The experiments show that the acoustic SLAM system based on the RIR cepstrum feature and the graph optimization method can accurately estimate the trajectory of the mobile terminal and the positions of doors and windows in the room map with an average self-localization error of2.78 cm and an average map building error of 4.13 cm.