Data fusion for obstacle identification using ultrasonic sensing

An in-situ, multi-level data fusion algorithm has been developed and implemented for an ultrasonic sensor-array that is embedded into a long cane as a travel aid for the blind. The algorithm is compact and efficient, appropriate for microcontroller-based portable applications. In this paper, several issues of the algorithm development are investigated, including echo signal quality verification and obstacle shape identification. A new technique based on dual-frequency ultrasound scanning is proposed. The algorithm development is based on an obstacle characterization scheme that describes obstacles using three representative groups. Analytical models for each group have been established