Gesture Recognition using a Neuro-Fuzzy Predictor

G. Bailador, G. Triviño, and S. Guadarrama (Spain)


Gesture recognition, prediction, neuro-fuzzy, accelerome ter.


Gesture recognition opens a wide range of possibilities for human computer interface development. This paper de scribes the design and the construction of a system for gesture recognition that includes hardware and software. The hardware part consists in a device that captures the acceleration provoked by the operator hand movement in the three axes. This device has a wireless connection, is powered by a long-lasting battery and it is small and light enough to be user friendly. The software part is built over an Adaptive Network Based Fuzzy Inference System (ANFIS) as a neuro-fuzzy predictor. Using this basic component as building brick, the gesture recognizer con sists of a structure of independently trained gesture pre dictors. The partial results of these gesture predictors are fused to build a robust gesture classifier. A remarkable characteristic of the developed system is its versatility and robustness supported by the fact that neither filtering nor normalization has been used. The results of several ex periments demonstrate the suitability of the presented approach.

Important Links:

Go Back