Study on object shape recognition using an ultrasonic sensor

An object shape recognizing method which analyzes an ultrasonic waveform is proposed. A non-directional pulse wave transmitted by an electric spark is employed. The experimental and theoretical waveforms reflected from severa l objects coincide well with each other and it is proved that each waveform has characteristics based on the surface shape of the reflected object. Th e pattern matching method, which compares a waveform in time domain directl y with registered templates of waveforms and uses a cross-correlation index , is proposed. This method is advantageous compared with that using a power spectrum, since the frequency conversion procedure is not needed. An exper iment of various objects recognition is carried out. The probability of rec ognition using a cross-correlation index is better than 98%. This method ca n cope with the change of the distance and the orientation of the object pr ovided that the templates of corresponding distances and orientations are p repared in advance.