Background
With advance of technology, context recognition is getting more and more attention, in particular, for IoT-deployed smart home services. However, existing works for context recognition are impractical and limited since they require additional infrastructures or multiple wearable devices. We propose a collaborative sensing framework to recognize user activities and locations in a smart home environment using commercial smart devices. We analyze multiple sensor data from a smartwatch and Wi-Fi signals collected from a smartphone.
Activity Fingerprinting
Smartphone cannot be attached to users when they are in home. So we use smartwatch, Samsung Gear S, for recognizing users activities. Its always attached to users and has various sensors, e.g. accelerometer, gyroscope, microphone, etc. Also, for increasing recognition accuracy, we use those sensors collaborativeellly.
Location Fingerprinting
Traditionally, many researchers used Wi-Fi signal to recognize users location. But there are some limitations in previous works, so we tried to address some of them, e.g. fluctuations of AP signal strength as time change.
Bluetooth Fingerprinting
Bluetooth signal is useful for determining whether the user and the smartphone are in the same room. We estimate the distance between smartwatch and the smartphone gathering the Bluetooth signal strength as Received Signal Strength Indicator. We use the distance information for correcting the location recognition.
Relevant Publications
- Jonghoon Shin, Hyunchoong Kim, Dayoung Lee, Yohan Ko, Seong-il Hahm, TaeJun Kwon, Kyoungwoo Lee, “Enhanced Indoor Localization in Home Environments Using Appearance Frequency Information”, IEEE International Conference on Systems, Man, and Cybernetics (SMC)
- Hyunchoong Kim, Jonghoon Shin, Soohwan Kim, Yohan Ko, Kyoungwoo Lee, Hojung Cha, Sungil Ham, Taejoon Kwon, “Collaborative classification for daily activity recognition with a smartwatch”, IEEE International Conference on Systems, Man, and Cybernetics (SMC), Oct, 2016.