Input action classification in a 3D gesture interface for mobile devices

In this research we propose a new motion classification method to improve operability of a 3D gesture interface that assists text input on mobile devices. A certain range of time-series finger scale data is cropped and is classified using linear discriminant analysis. To confirm possibility of linear separation, data were visualized using principle component analysis. Experimental result with changing cropping ranges and sampling rates showed that the recognition rate improved when the cropped time is longer, and more than 97.9% recognition rates were achieved with cropping time of 0.77s/0.38s from both/one sides of the peak.