| Summary: | This thesis presents the development of tangible acoustic interfaces for human computer interaction. The method adopted was to position sensors on the surface of a solid object to detect acoustic waves generated during an interaction, process the sensor signals and estimate either the location of a discrete impact or the trajectory of a moving point of contact on the surface. Higher accuracy and reliability were achieved by employing sensor fusion to combine the information collected from redundant sensors electively positioned on the solid object. Two different localisation approaches are proposed in the thesis. The learning-based approach is employed to detect discrete impact positions. With this approach, a signature vector representation of time-series patterns from a single sensor is matched with database signatures for known impact locations. For improved reliability, a criterion is proposed to extract the location signature from two vectors. The other approach is based on the Time Difference of Arrival (TDOA) of a source signal captured by a spatially distributed array of sensors. Enhanced positioning algorithms that consider near-field scenario, dispersion, optimisation and filtration are proposed to tackle the problems of passive acoustic localisation in solid objects. A computationally efficient algorithm for tracking a continuously moving source is presented. Spatial filtering of the estimated trajectory has been performed using Kalman filtering with automated initialisation.
|
|---|
