Sonar - now a viable alternative to vision

Citation

Kay, L. (1999), "Sonar - now a viable alternative to vision", Sensor Review, Vol. 19 No. 3. https://doi.org/10.1108/sr.1999.08719caa.002

Sonar - now a viable alternative to vision

Sonar - now a viable alternative to vision

The authorLeslie Kay is at Spatial Sensing Laboratory, Bay Advanced Technologies Ltd, Russell, New Zealand.

Keywords Sonar, Ultrasonic, Vision

There is now considerable interest in the use of ultrasonic technology as a means for spatial sensing as an alternative to optical sensing. The papers in this special issue on ultrasonic sensing are front runners in the search for an effective robot sensor. They are using evolving technology. Only two papers however employ fully customised appropriately designed systems based on the optimum use of physics of sound, just as is commonly done in underwater sound and body imaging. The restrictive elements are the transducers forming the sensing array. These are designed optimally to match the signal processing needed to gather the desired information. All other systems use commercially available transducers designed for a specific purpose in the early 1960s. These systems suffer from the serious limitation of making the signal processing match the transducer characteristics. This never happened in the development of underwater sensors for the location and classification of objects of interest - fish and marine mammals, submarines, mines and torpedoes. The transducers were specially developed for the optimal gathering of the desired information that matched the available information-processing technology. This must be more so than in the 1950s when I was deeply involved in the research and development of military sonars.

The Sonic Torch of the 1960s from which the Polaroid transducer was legitimaly copied, the Sonar Glasses of the 1970s and the Trisensor of the 1980s used transducers of varying shape, size and frequency response. They were specially designed and developed to meet the information gathering needs of a sensor for blind persons as new knowledge and understanding of the problem evolved. Blind persons became more able to utilise the new spatial information as it became available.

The difficulties experienced in developing a transducer technology were considerable and needed appropriate experience and special craft skills. We gradually acquired this at Canterbury. Very importantly, display of the information that came available required real-time spectrum analysis which was expensive. A bulky bank of very narrow bandwidth filters with a fast scanning switch was needed in the early days. Alternatively one could use one's ears as was done in military sonars and as is done in the sensors for the blind. By this means we found that the rich sensory information needed was evidently available. Blind persons were found to be able to discriminate between a smooth pole and a tree trunk because of the roughness of the bark. They could even discriminate between a shrub with small leaves and a shrub with large leaves. They could recognise their surroundings and map them.

This ability by a machine has now been found to be possible using a computer analyser as an alternative to the auditory sense. High speed high quality FFT analysis can now be downloaded from the Internet eliminating the need for special computer hardware.

There is a further reason why even the Polaroid transducer is not optimally used. The use of pulse-echo techniques has led transducer designers to test transducers using impulses. This highlights the variable sensitivity and frequency response of the transducer. A specified peak frequency has been sought by designers but not found. If instead the overall sensitivity and frequency response is kept within reasonable tolerances, near optimum performance can be obtained driving the transducer in a linear sawtooth over the desired frequency band.

Greatly improved signal to noise performance is then possible for a specified object surface, loosely called a target in the literature. Integration of the signal over a sweep period produces the desired matched filter performance as described in the two papers.