Think of smart home, think futuristic. For instance, imagine a scenario where your home is so smart that it completes sentences for you. Okay, not that futuristic. But this home will complete your tasks, for sure.
A researcher at the IITB-Monash Research Academy, Mumbai, is writing algorithms to make sure that your home does just that.
Sequence labelling
Naveen Nair, who started as an employee of the health department in Kerala after a cursory education, is now working on his Ph. D at the Academy. His passion, sequence labelling, led him to his research project and the vast practical applications that it held out.
Smart homes, or homes wherein appliances are turned on or off with direction from us — either pre-loaded cues, or through speech recognition — already exist. Apparently, these technologies are based on a process called sequence labelling, which classifies how tasks occur in a consequential sequence.
To help senior citizens
Naveen says his research, when it steps out of the lab, has the great potential to make things comfortable for senior citizens living alone, and also allow remote monitoring of their health and safety.
Naveen attempts to take the existing models of sequence labelling to the next level. According to him, his attempt is to study the relationship among different actions of an individual, and from different inputs on the same sequence of actions.
For instance, he says, if a person goes to the microwave oven to heat his food, picks a glass of water from the fridge and walks back to sit at the dining table, if these three objects can speak to each other and used as signals for the next action in sequence, say, turn on the dining room light. It could also provide inferences as to the state of health of the individual by recording if he or she is carrying on routine activities.
Naveen’s system automatically learns such relationships in an efficient and optimal way. This information could be used to optimise the way devices are utilised in a smart home. Whereby, when a person opens (and closes) their bedroom door and turns off the bedside heater, the system infers that the person is going to sleep and automatically switches off the television in the living room.
Improves accuracy
The research will have a variety of other practical applications. “As my research is based on learning optimum features for structured output, it will improve accuracy and can be used to perform efficient activity recognition,” he says.
It may also aid systems that optimise device utilisation, and enable power saving.
“Other potential applications for this technology may be as broad as in named entity recognition whereby more accurate identification of words within text are then used to extract journalistic articles,” he explains.
Practical solutions
CEO of IITB-Monash Research Academy Mohan Krishnamoorthy says Naveen’s research surely advances and takes ahead of sequence labelling. But the nub is in translating research into practical solutions and taking Naveen’s algorithms to read and process information from sensors placed in actual homes would be the ultimate test.
Naveen’’s research is under the guidance of Professors Ganesh Ramakrishnan (the Indian Institute of Technology, Bombay) and Shonali Krishnaswamy (Monash University, Australia).
