Accelerometer-Based Gesture Recognition with the iPhone - Marco Klingmann, Msc Cognitive Computing

The growing number of small sensors built into consumer electronic devices, such as mobile phones, allow experiments with alternative interaction methods in favour of more physical, intuitive and pervasive human computer interaction. This research examines hand gestures as an alternative or supplementary input modality for mobile devices. The iPhone is chosen as sensing and processing device. Based on its built-in accelerometer, hand movements are detected and classified into previously trained gestures. A software library for accelerometer-based gesture recognition and a demonstration iPhone application have been developed. The system allows the training and recognition of free-from hand gestures. Discrete hidden Markov models form the core part of the gesture recognition apparatus. Five test gestures have been defined and used to evaluate the performance of the application. The evaluation shows that with 10 training repetitions, an average recognition rate of over 90 percent can be achieved.