Smartwatches and other wearable devices have become essential data acquisition platforms for capturing user states and environmental information in real time. Wear OS based smartwatches, in particular, provide a variety of sensors, including accelerom...
Smartwatches and other wearable devices have become essential data acquisition platforms for capturing user states and environmental information in real time. Wear OS based smartwatches, in particular, provide a variety of sensors, including accelerometers, gyroscopes, heart rate sensors, and ambient light sensors, that can be utilized in diverse application domains such as activity recognition, healthcare monitoring, and context aware services. However, continuously collecting and transmitting high frequency sensor data to external devices introduces challenges related to power consumption, packet loss, transmission latency, and platform dependency. Furthermore, existing Wear OS sensor collection approaches are primarily designed for integration with Android smartphones, making it difficult to directly utilize sensor data in Windows or Linux PC based external systems.
In this study, we design and implement a cross-platform sensor integration toolkit that enables multi-sensor data collected from Wear OS smartwatches to be utilized not only on Android smartphones but also on Windows and Linux PC based receivers. The proposed toolkit allows users to directly configure collection conditions, including the types of target sensors, the sampling period, and the transmission method. To handle sensors with different sampling characteristics, such as accelerometers, gyroscopes, heart rate sensors, and ambient light sensors, the toolkit defines a common data frame at the sensor event level instead of recording all sensor values in a single row, thereby preserving each sensor's inherent sampling characteristics while enabling the data to be interpreted in the same manner across diverse receiver environments.
For data transmission, the proposed toolkit provides two complementary transmission paths: a Data Layer API based path and a BLE based path. For the BLE path, we adopt a Hybrid structure that combines Advertising and GATT. Advertising is used as a lightweight connectionless path that allows multiple receivers to observe sensor data and assists in establishing GATT connections, while GATT serves as a dedicated path for continuous sensor data streaming with a selected receiver. In addition, we define a common data frame containing Seq and Watch_TS fields to enable quantitative analysis of packet loss rate, reception interval, and transmission latency.
Furthermore, based on the user defined sensor types, sampling period, transmission method, and CSV storage options, the toolkit provides a feature that automatically generates code snippets for sensor collection, frame encoding, transmission and reception, and CSV storage. This reduces the repetitive implementation burden and allows the toolkit to be used as a reusable research and development support tool for developing wearable sensor based application services.
To evaluate the proposed toolkit, we collected accelerometer data from a Wear OS smartwatch and compared the reception characteristics of Data Layer API, BLE GATT, and BLE Advertising across Android smartphones, Windows PC, Linux PC, and Android tablets. The results show that on Android smartphones, Data Layer API achieved an average loss rate of 3.19%, GATT achieved 4.82%, while Advertising exhibited a substantially higher loss rate of 80.58%. In Windows and Linux PC based environments, GATT demonstrated stable reception with a 0% Seq based loss rate over 30 minute long duration experiments. In contrast, Advertising showed significant variation depending on the receiver environment, with high loss rates on Android smartphones, Windows PC, and tablets, but relatively low loss rates on Linux PC. These findings indicate that, rather than being used as a high reliability long duration streaming path, Advertising is better suited as an initial reception and GATT connection assistance path, while GATT serves as the actual continuous streaming path. The proposed Hybrid structure is therefore well suited for cross-platform sensor data transmission in Wear OS based environments.