| Summary: | 碩士 === 國立臺北科技大學 === 自動化科技研究所 === 107 === How to effectively prevent and create a safe environment for home has become an important topic. However, different kinds of anti-theft devices can be found in the market. Although wireless anti-theft systems are prevalent in the market, there have been problems with high cost and power consumption. Besides, they are not intelligent to recognize different patterns of thief behaviors. To mitigate power consumption, this paper developed a home security anti-theft system using LoRa communication technology. To learn new patterns, we deployed a recurrent neural network (RNN) and Dynamic Time Warping (DTW) in the system. This system used a STM32L476 microcontroller, a LoRa wireless transmission module, and a 3-axis magnetometer to detect the status of doors and windows. The obtained magnetometer data was pre-processed by the filters. Then, learning RNN was used to train the system for any customized environment. The detection results were transmitted to the host via LoRa by a customized protocol. From the experimental results, it was verified that the LoRa low-power characteristics deployed in the microcontroller can achieve a high recognition rate of 95%, and can last 3300 hours of use with a 210mA button battery.
|
|---|
