Yingting Gao / 高莹婷
Sensing with OBean-M: Odor-related indoor activity recognition using a MEMS sensor array electronic nose
/ Introduction
​
Using olfactory-based sensing technology as an non-intrusive method to detect human activity is a new trend of ubiquitous computing area, which has unique advantages in healthcare, dietary management and home automation. We presented OBean-M, a novel electronic nose for detecting odor-related activities in the home environment using an array consist of 16 micro-electro-mechanical systems (MEMS) gas sensors. We evaluated OBean-M system for identifying 18 different daily object in a laboratory setting with an average F1 score of 0.99. Then we deployed OBean-M system in a real-world household. The results shownour system could accurately distinguish from 7 different activities with average F1 score of 0.96. Moreover, OBean-M also was proved the ability of distinguishing different sub-activities in some complex odor-related activities.
To have a better understanding, we described the odor-related human activity in a hierarchical structure. The structure consist of following three levels:
-
Object/source-level – The basic level of the odor-related human activity, stands for the odorous object to beinteracted with human.
-
Activity-level – The middle/top level(depends on the complexity of the activity) of the odor-related humanactivity, stands for the odor emanates from the interaction between the odor object and human.
-
Context-level – The top level of the odor-related human activity, stands for the context consist of multipleactivities
We deployed the OBean-M prototypes in seven odor ’hot zones’. And they were placed in positions suitable for receiving odor flow based on our evaluation experience.