A new approach of hybrid decision tree

Blind people need some aid to interact with their environment with more security, especially, to avoid collision and falling. The process of staircase negotiation is complex for visually impaired people. Therefore, an intelligent system is worth helping them. In this paper, we investigate incorporating only one ultrasonic sensor within an electronic white cane to recognize floor state and classify the environment in three classes, even surface, ascending stair case and descending stair case. In our knowledge, no previous work was concerned with such context. The performance of floor state recognition system depends, firstly, on signal processing strategy, then, on objects representation and, finally, on classification approach, making the decision appropriate. In this paper, the proposed strategy consists in merging more than one transformation of ultrasonic signals, mostly Fourier transform and wavelet transform, used as features in a hybrid decision tree.