Differentiation and classification of different materials on a touch-screen using a piezoelectric contact microphone

Abstract: In this report, a potential improvement of an optical touch screen solution by means of acoustic sensing is presented. The goal is to utilize the different vibration patterns that arise when different materials touch the screen to be able to identify them. This identification of different materials opens up new opportunities to give input to the touch-device which could potentially replace menu bars or touch-andhold interaction that exists today. The ability to discriminate between object also gives the opportunity for different objects to represent different things, such as colors, tools or unique users. A piezoelectric accelerometer is used to collect the sound signals, which are then classified using the K-Nearest-Neighbour method to identify which material that touched the screen. The result shows that the classification for three different objects was done with a 96% correct classification. If only two different pens are used, the correct classification goes above 99%. The report further describes some of the key factors that needs to be taken in account when developing such an acoustic sensing system, such as force, distance and location of the taps on the screen. Those factors does all seem to alter the frequency spectrum in some way.