Description
The recent advances in low-power wireless sensor-based technology have made it possible to design a variety of wearable devices for applications such as environment sensing, human activity monitoring and tracking, etc. These wearable devices have a variety of embedded sensors, which have the capability to collect multiple streams of data. These data can be transmitted wirelessly to a cloud server or logged locally for data-based applications. The devices are widely used in scientific research. Such data collection capabilities make wearable devices a very cheap and accessible solution that fits into people’s daily routines, so their use has expanded beyond remote-monitoring applications to tracking military personnel or animals etc., since human intervention can be very limited. In this thesis, we developed a functional wearable device capable of tracking someone’s location and activity based on Bluetooth Low Energy (BLE) wireless technology. A network of these devices can collect research data in applications such as tracking an individual’s geographic location, tracking social interactions, determining situational behavior etc. Thus, collecting data are the most important function of a wearable device. Low-power sensors embedded in this device can collect a large volume of data with a very high resolution or frequency. Since these devices are typically powered by batteries, they have a limited power budget for transmitting, receiving, and logging data, or for performing computations. In our wearable device, the data are logged locally on a micro SD card. We observed that data collection accounted for a significant proportion of the power consumed by our wearable device, especially the micro SD write operations. We implemented two different software-based data-reduction techniques: batch writes and data compression algorithm. These techniques reduced the size of the sensed data and thus, contributed to reducing the average power consumption of the micro SD write operations. The batch writes technique reduced the average power consumed by the micro SD module by 74% and the compression algorithm reduced it by 80%. Finally, we simulated these results for a custom designed system, based on a similar platform, to determine the effectiveness of smart data writing methods in application specific hardware.
