基于人类回声定位的目标识别方法研究

发布时间：2018-08-02 12:36

【摘要】：回声定位是人类的一项基本的认知技能。对回声定位的研究具有重要的研究意义和应用价值。通过对人类回声定位机理的研究,可以提炼出一种新的面向环境感知和自主导航的信号处理方法。一方面,可为制造盲用的便携式回声定位装置提供一定的理论依据和技术支撑;另一方面,将这种认知方法引入雷达、声呐等传感设备中,可有望为其提供一种全新的思路和信号处理方法。本文在对面向二维形状目标识别的人类回声定位行为实验进行深入研究和分析的基础上,搭建了一套人类回声定位实验系统,利用该系统模拟了人类回声定位行为实验中的两个场景:“保持静止”场景和“自由运动”场景,分别在两个场景下进行了信号采集工作,研究了相应的声音信号特征提取方法,并分别进行了两个场景下的二维形状目标识别方法研究。本文的主要工作可总结如下:1)对面向二维形状目标识别的人类回声定位行为实验的结果进行了研究和分析,搭建了与之对应的面向人类回声定位的实验系统,并进行了相应的测试;2)研究了声音信号特征提取方法,提取出了总能量(TS)、平均频率(AF)、低频能量(LFS)、中频能量(IFS)、高频能量(HFS)、亮度(TB)、三色值(T1、T2、T3)、不规则度(IRG)等10个特征;3)模拟了“保持静止”场景下的人类回声定位实验,分别从采集的原始声音信号和特征提取两个方面进行了分析,结果表明:很难通过单一位置的声音信号区分不同形状目标,该结果与行为实验的结果相符;4)模拟了“自由运动”场景下的人类回声定位实验,分别对40cm复合信号、80cm复合信号以及80cm回声信号进行了特征提取和可视化分析,结果表明:根据这三种声音信号的总能量特征或者平均频率特征均可有效区分不同形状的二维目标,该结果与行为实验的结果相一致。
[Abstract]:Echolocation is a basic cognitive skill. The study of echolocation has important significance and application value. By studying the mechanism of human echolocation, a new signal processing method for environment perception and autonomous navigation can be extracted. On the one hand, it can provide some theoretical basis and technical support for the manufacture of portable echolocation devices for blind use; on the other hand, it can be introduced into radar, sonar and other sensing devices. It is expected to provide a new way of thinking and signal processing. Based on the deep research and analysis of human echolocation behavior experiment for two-dimensional shape target recognition, a set of human echolocation experiment system is built in this paper. The system is used to simulate two scenarios in the human echolocation behavior experiment: "keep still" scene and "free motion" scene. The corresponding feature extraction method of sound signal is studied, and the two dimensional shape target recognition method under two scenes is studied respectively. The main work of this paper can be summarized as follows: (1) the experimental results of human echolocation behavior for two-dimensional shape target recognition are studied and analyzed, and a corresponding experimental system for human echolocation is set up. A corresponding test is carried out to study the method of feature extraction of sound signal. Total energy (TS), average frequency (AF), low frequency energy (LFS), intermediate frequency energy (IFS), high frequency energy (HFS), brightness (TB), tricolor (T _ 1 T _ 2 T _ 3), irregular degree (IRG) and so on 10 characteristics are extracted to simulate human echolocation experiments under "keeping still" scene. The results show that it is difficult to distinguish different shape objects by single position sound signal. The results are in agreement with the results of behavioral experiments. (4) the experiments of human echolocation in the "free motion" scene are simulated. The feature extraction and visual analysis of the 40cm composite signals and the 80cm echo signals are carried out, respectively. The results show that the two dimensional targets with different shapes can be effectively distinguished according to the total energy characteristics or the average frequency characteristics of the three kinds of sound signals. The results are in agreement with the experimental results.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN912.3

【参考文献】