An Exploratory Study of Simple Fall and Activity Recognition Using mmWave

Sammanfattning: Background. As smart appliances become more attractive, the demand from public consumers grows, and producers are in search of innovative technologies that may aid in the creation of smart homes. Current products may use screens and buttons, voice commands and motion detection to create an interactive experience for consumers. A rather new technology that has gathered attention in recent years is millimetre wave radar sensors (mmWave). This technology uses electromagnetic waves to detect objects in the vicinity of the physical sensor; it may detect both the range, velocity and orientation of an object in relation to itself. The current research has had a main focus in automotive and industrial industries, and the technology has thus far been applied to areas such as vital signs monitoring, people counting, motion control, object detection and collision aversion among others. An attractive feature for use in smart homes that the sensor provides, or rather lacks, is its inability to identify different people. As the information gathered is a point cloud -- in low resolution -- any monitored people retain their privacy under normal circumstances. Objectives. The aim of this thesis is to verify the usability of mmWave sensors in smart homes, as well as reaching an initial understanding of people's opinions regarding the mmWave technology. Method. Experiments are performed to test how well the mmWave sensors can determine if a person is standing, sitting, lying or if they have fallen. The approach for the developed program to make these predictions are done through simple algorithms. Experiments were performed in an environment that was meant to mimic the conditions of a home. Participants were also asked about their opinion of potentially using the technology in their home, both regarding imagined usage and whether the sensor would cause them any discomfort. Results. The results show that while the implemented software in this thesis helps validate the proof of concept for the intended purpose, the technology shows a lot of promise for the future. Further algorithmic efforts will however be required to reach the desired maturity. The opinions of the participant show a generally positive response in using the sensor, however, they also note that if the sensor is to be used in their home, any data gathered should be both available and in control of the consumer to ease suspicions of misuse. Conclusions. The authors conclude that while not yet quite ready, the sensor is indeed a probable candidate to be integrated into smart homes of the future.