Artiicial Neural Networks for Tyre Pressure Estimation

We present an application of artiicial neural networks to monitoring the tyre pressure of cars using acoustic data derived from vibrations recorded by an accelerometer. As a preprocessing step for feature extraction we perform a spectral analysis of the sound data. The components of the estimated power density spectrum are used as input patterns for the neural estimator. These patterns are used to train a neural estimator for the tyre pressure. We compare the designated neural structure to a common least mean square approach based on a linear model and show, that the continuous function approximation network performs better in a least square sense than the linear model.