Master Theses - Details
Novel User Interfaces for Hearables
Area: User Interface Design, Machine Learning
Tools: Matlab, Python, Estimation Theory, Machine Learning
Categories: Bachelor Thesis, Master Thesis
In recent years, wireless bluetooth earbuds finally arrived at market for consumer electronics. These so called hearables have the potential to improve the hearing and listening experience of the wearer. The possible features include selective listening, augmented reality, active noise cancellation and much more.
However, the use of these features is still limited, in part because there is no suitable user interface available. Current devices mostly employ physical buttons or touchpads. Often the number of buttons / touchpads is low, and thus, the number of possible commands. This limits the possible control of more complex features by the user. Additionally, they cannot be used handsfree, are often limited in robustness, and produce undesired sounds in the ear canal due to the occlusion effect. Novel interfaces could thus improve the user experience of hearables substantially.
The novel interfaces would need to accurately and robustly detect commands issued by the user, while being as unobtrusive as possible. A number of things could function as a command, for example: head gestures, or sounds created by the user. A first step in this thesis would be analyzing different kinds of commands for their suitability as an interface. This involves evaluating if the commands can be detected easily and robustly. In order to achieve this, corresponding signals need to be recorded and suitable detection algorithms implemented. This includes usage of machine learning methods. The implementation should reflect the importance of computational constraints of the hearable platform. The implementations shall be evaluated based on suitable objective criteria, such as false positive / negative rate. Additionally, the subjective performance of the interfaces shall be evaluated by informal usage tests.