VoiSense
Sirat Samyoun, John A. Stankovic · 2021
Data collection is a major hurdle faced by the researchers, particularly in the field of wearable computing, as it requires significant amount of time and manual effort. Although wearable devices such as, smartwatches have sophisticated sensing capabilities today and are being popularly used by CPS researchers, collecting data from the sensors embedded on the smartwatch remains a challenging job. The fundamental limitation of the smartwatches is the small form factor, which makes it difficult for the user to input any data collection preferences or labels for the data collection. The existing apps/systems that takes such inputs manually from the user are therefore ineffective for the smartwatches. In this paper, we present the idea of using voice interaction capabilities of a smartwatch for conveniently collecting and labeling sensor data from the available sensing modalities. We present VoiSense, a voice interactive smartwatch based system which intelligently gathers the user preferences and data labels with voice interaction and facilitates convenient data collection without requiring any expertise or significant manual effort.