Abstract
Smart home technologies could play an important role in enhancing the lives of the elderly from health monitoring through to improved adaptive heating con
trol. Such technologies can benefit from incorporating information pertaining to building utilization if it were quantified accurately. Domestic occupancy sensingbased on passive infrared technology relies on continuous movement to detect room usage, however, withoccupants of advancing years, reduced levels of act ivity may lead to false negatives being incurred through this approach. More accurate occupancy sensing would enhance benefits enabling smarter monitoring of health and well-being, as well as providing useful input to zonal or community heating systems to allow them to schedule energy usage more efficiently. If occupancy sensor technology were to underperform it could be critical to the vulnerability of residents of sheltered accommodation, highlighting the need for robust, consistent detection of living space usage.
The study presented here develops an algorithmic approach to deliberating between a committee of multiple low cost sensors deployed in residential premises to determine the state of occupancy in a given room. The approach combines logic with machine learning to classify changes in sensor measurement patterns during room entry and exit events to infer the true
occupancy state. The data collection infrastructure designed and used in field trials is described along with the methodology used to benchmark candidate model performance. The resulting model requires minimal computational effort and can also take advantage of any pre-existing sensors. The first 2 weeks of
experimental data is presented, demonstrating improvement in the accuracy of occupancy detection,particularly for single elderly people, over PIR based systems achieved through intelligent analysis of data.
trol. Such technologies can benefit from incorporating information pertaining to building utilization if it were quantified accurately. Domestic occupancy sensingbased on passive infrared technology relies on continuous movement to detect room usage, however, withoccupants of advancing years, reduced levels of act ivity may lead to false negatives being incurred through this approach. More accurate occupancy sensing would enhance benefits enabling smarter monitoring of health and well-being, as well as providing useful input to zonal or community heating systems to allow them to schedule energy usage more efficiently. If occupancy sensor technology were to underperform it could be critical to the vulnerability of residents of sheltered accommodation, highlighting the need for robust, consistent detection of living space usage.
The study presented here develops an algorithmic approach to deliberating between a committee of multiple low cost sensors deployed in residential premises to determine the state of occupancy in a given room. The approach combines logic with machine learning to classify changes in sensor measurement patterns during room entry and exit events to infer the true
occupancy state. The data collection infrastructure designed and used in field trials is described along with the methodology used to benchmark candidate model performance. The resulting model requires minimal computational effort and can also take advantage of any pre-existing sensors. The first 2 weeks of
experimental data is presented, demonstrating improvement in the accuracy of occupancy detection,particularly for single elderly people, over PIR based systems achieved through intelligent analysis of data.
Original language | English |
---|---|
Publication status | Published - Aug 2014 |
Event | OB-14 - Nottingham, United Kingdom Duration: 5 Aug 2014 → 6 Aug 2014 |
Conference
Conference | OB-14 |
---|---|
Country/Territory | United Kingdom |
City | Nottingham |
Period | 5/08/14 → 6/08/14 |
Keywords
- Elderly people
- Occupancy Sensing