Fractal Brownian motion for feature extraction in noisy signals from gas sensors

The present work reports on a new pattern recognition method applied to the electrical response of a gas sensor. The sensor is a tin oxide device used as an electronic nose. It yields noisy responses when submitted to organic solvents. Signals were analyzed by the epsilon-blanket fractal dimension associated with the fractal Brownian motion. The classification of features shows close to 100% recognition rate for five different gases.