基于GFCC特征提取机器人仿生声呐SLAM算法研究

Research on SLAM algorithm of robot bionic sonar based on GFCC feature extraction

针对移动机器人在室内环境下的定位问题,本文采用基于GFCC((Gammatone Frequency Cepstrum Coeffi-cient))特征提取BatSLAM模型,用指数压缩来模拟听觉系统的非线性特性,使用加海宁窗处理来减小回波信号所存在的边缘影响,使用离散余弦变换来对耳蜗图进行有损数据压缩,从而提高耳蜗图的抗干扰能力,使用升半正弦倒谱提升来提高耳蜗图的鲁棒性,通过GFCC特征提取可以有效提高室内定位的精度和准确性。实验表明,基于GFCC特征提取Bat-SLAM模型,通过提高耳蜗图的抗干扰性和鲁棒性,可以有效的较小定位误差,从而提高移动机器人的定位精度和准确性。

In view of the positioning problem of mobile robot in indoor environment,based on the traditional RatSLAM model,using sonar sensor instead of visual sensor GFCC feature extractionl,BatSLAM mode can meet the construction of two-dimensional experience map in dark light environment.BatSLAM model uses Gammatone auditory filter banks to filter,but its robustness is very poor in noise environment,which is easy to be affected by external noise,which will lead to great differences in cochlear map at the same position.The positioning accuracy and accuracy are poor.The BatSLAM model based on GFCC fea-ture extraction is used to simulate the nonlinear characteristics of auditory system by exponential compression.The edge effect of echo signal is reduced by adding Haining window processing.Discrete cosine transform is used to compress cochlear map with lossy data,so as to improve the anti-interference ability of cochlear map.The experiment shows that the BatSLAM model based on Gammatone Frequency Cepstrum Coefficient can effectively reduce the positioning error by improving the anti-inter-ference and robustness of cochlear map,thus improving the positioning accuracy and accuracy of mobile robot.