This paper proposes a developer-friendly multi-modal sensing toolkit designed to support the efficient configuration and rapid development of sensor-based wearable applications on standard Wear OS smartwatches. The proposed toolkit, named MST-Wear OS, provides a comprehensive framework for multi-sensor data collection, empirical comparative testing of sensor sampling rates, monitoring of battery consumption usage, and exporting automatically generated Kotlin code snippets with JSON data files. From a companion mobile application user interface, developers can easily configure various parameters, such as sensor type, sampling interval, and collection duration. These settings are then delivered to a paired Wear OS smartwatch via the Data Layer API. The specified sensors on the smartwatch are activated to collect sensor data based on the received settings. This data is saved as JSON files and transferred to the companion mobile application for further analysis. This toolkit enables developers to explore and compare sensor configurations effectively, significantly enhancing their productivity when developing sensor-based machine learning Wear OS applications. We demonstrate the feasibility and usefulness of this toolkit with two prototype application scenarios and configurations.
Park et al. (Fri,) studied this question.