期刊名称：Inteligencia Artificial : Ibero-American Journal of Artificial Intelligence
印刷版ISSN：1137-3601
电子版ISSN：1988-3064
出版年度：2008
卷号：12
期号：37
页码：27-38
出版社：Spanish Association for Intelligence Artificial
摘要：On studying some of principal works about objects basic shape recognition in a structured environment, we can conclude that most of them use time characteristics from captured echoes. Altohug frequency analysis has been useful in audio signal processing, it has no mean the same role in ultrasonic signals applied to robotics because the use of reduced wideband transducers. Moreover, the Wavelet Transform, as a tool for analyzing frequency components in time space, has no been used widely in ultrasonics. Along this article, a system of ultrasonic signal processing, is proposed through a 2 Transmitters ? 2 Receptors sensorial structure, in such a way it is able to recognize walls, corners, edges and cylinders in only one exploration with a high success percentage, measuring different characteristics from time space, Fourier frequency spectrum and Wavelet coefficients, selecting the best trough Principal Component Analysis (PCA), and using them for training specialized Artificial Neural Networks (ARN) as classifier.On studying some of principal works about objects basic shape recognition in a structured environment, we can conclude that most of them use time characteristics from captured echoes. Altohug frequency analysis has been useful in audio signal processing, it has no mean the same role in ultrasonic signals applied to robotics because the use of reduced wideband transducers. Moreover, the Wavelet Transform, as a tool for analyzing frequency components in time space, has no been used widely in ultrasonics. Along this article, a system of ultrasonic signal processing, is proposed through a 2 Transmitters ? 2 Receptors sensorial structure, in such a way it is able to recognize walls, corners, edges and cylinders in only one exploration with a high success percentage, measuring different characteristics from time space, Fourier frequency spectrum and Wavelet coefficients, selecting the best trough Principal Component Analysis (PCA), and using them for training specialized Artificial Neural Networks (ARN) as classifier.
