What makes connected homes smart? First and foremost, it’s intelligent processing. Second, it’s automation of previously complex manual configuration. Thanks to machine learning, smart home manufacturers can solve both problems. Or can they?
How do Smart Home Systems Powered by Machine Learning Work?
The devices which comprise an AI-powered Home Automation system should be able to provide services both to each other and to the end user – and continuously acquire new data for further learning.
The key part of such solutions constitutes a data analytics module where raw sensor readings are processed and data patterns are identified, so the system knows event A (a home owner unlocks the front door) is usually followed by event B (the heating is turned up).
With those data patterns, a Smart Home system should be able to predict a user’s behavior based on historical data and develop the so-called situational awareness – i.e., understand a user’s intentions at a given moment and change parameters accordingly.
In an ideal world, it may look like this:
During the weekdays, John’s alarm clock goes off at 6 am. His Smart Home system automatically puts the blinds up and sends a notification to the connected coffee maker. Having taken shower, John goes to the kitchen where a cup of hot cappuccino is waiting for him (that’s right, the system knows he prefers cappuccino in the morning). On Sundays, however, the connected home system does not do all that till 10 am, because John likes to have a lie-in on weekends.
Confusion over What Smart Home Really Is
A Nest thermostat is an obvious example of a self-learning connected home solution. But is it really a Smart Home device? It is just an autonomous Wi-Fi-enabled gadget which uses temperature, motion, humidity and light sensor data to create custom temperature schedules, thus saving users energy and keeping them comfortable.
Are Self-learning Connected Home Solutions Actually Smart Home?
A truly intelligent Smart Home is a multi-layer system which requires little to no management on a user’s part and is capable of making decisions based on historical and real-time data.
Thus, the system should be able to identify significant user actions (Dave’s woken up), assess the probability events those actions trigger (Dave will now go to the bathroom) and issue appropriate commands to other devices within the network (turn on the lights in the bathroom).
The goal of Home Automation is to bring down any manual settings to zero. Currently, many Smart Home offerings are missing this. However, using machine learning can significantly minimize the inconvenience for connected home owners, who often have to set up and operate their not-so-smart devices manually.
Machine Learning Applications in Home Automation
Although Smart Home solutions aren’t likely to master context-based decision making in the foreseeable future, machine learning can make connected homes a lot smarter.
Face Recognition
Many Smart Home security systems incorporate the face recognition technology into connected video cameras. The neural networks identify facial landmarks – for instance, eyes, cheekbones, nose and chin – in a person’s photo and compare the data to the imagery produced by cameras. An AI-based security system can successfully identify the faces of a home’s residents and send notifications about suspicious activity to a home owner’s smartphone.
Biometric Access Control
Biometric door locks such as August, Kwikset (News - Alert) Kevo and Samsung can be seamlessly integrated into Amazon, Google or Samsung Smart Home ecosystems. They usually take photos of a user’s fingertips via optical scanners and store them for matching (once again, a computer vision is required) or use two-factor authentication (fingerprints and passwords).
Natural Language Processing
Home Automation solutions in many cases will rely on voice recognition technology and NLP tools like Amazon Transcribe and Azure Custom Speech Service or intelligent personal assistants like Siri, Alexa or Google (News - Alert) Home. The system isolates a person’s voice from background noises, converts the audio to a digital file and sends it to the cloud for NLP analysis. The cloud-based server then mines the meaning from other resources, as well as the Smart Home’s own database, and triggers an appropriate action.
The Challenge Is in Smart Home Market Fragmentation
In order to set up a secure, scalable and highly functional connected home solution, a user has to connect gadgets created by different vendors – and few Smart Home solutions are designed to communicate with 3rd-party products.
Over the last couple of years a few steps were made to solve the Smart Home interoperability issue. One of them is Z-Wave. It allows connected home manufacturers to integrate their devices with quite an impressive number of Smart Home solutions. Still, a random Wi-Fi switch in your bedroom won’t persuade your coffee maker to brew a cup of Americano – unless we’re talking about a combination of a simple coffee maker with WeMo and Alexa.
About the author: Andrei Klubnikin is Senior Content Manager at R-Style Lab – IoT software development company with a business office in San Francisco, California. Andrei has been a tech blogger since 2011 and currently writes for Clutch.co, Business.com, StartUs, GameAnalytics, Mobile App Daily, SmallBizClub.
