Getting Started with Voice-controlled Home Automation: Tips for IoT Start-ups

Consumers increasingly invest in the Home Automation technology to achieve multiple goals: optimize energy and water consumption, manage emergency systems remotely, orchestrate home appliances through a single touchpoint and access entertainment content in a hands-free mode. Most Smart Home systems, however, simply pair a connected device to a stationary control device and a custom mobile app which require users to manage the solutions manually, albeit remotely. Considering the fact 47.3 million Americans now have a smart speaker, it’s true to say voice interfaces are the future of Smart Homes. In this article, we’ll provide several tips for building a custom voice-controlled Home Automation solution.

Unlocking the Power of the Voice Recognition Technology

Voice-controlled Home Automation solutions allow users to manage home appliances (including non-electronic tools) via voice commands. A voice capturing device – for example, a smart speaker like Echo or Elechouse voice recognition module attached to an Arduino board – is the key component of such systems, although you need a display-based device to adjust a connected home’s settings; again, it can be a smartphone , tablet, custom control device or speaker with a screen.
An automated speech recognition (ASR) program lies in the heart of a voice-activated Home Automation solution. Such programs are powered by the voice recognition technology which captures spoken words, digitalizes sound by taking precise measurements of a sound wave, removes background noise and passes the data for further analysis. During the process, digital signals are broken down into smaller samples; their length depends on the duration of a sound produced by human vocal cords. Then the program compares these patterns with the phonemes of a target language and deciphers what a user has said. Based on this information, voice assistants can generate relevant replies and send commands to other connected gadgets within a Smart Home ecosystem.
The groundbreaking technology, in its turn, is based on artificial neural networks – algorithms that take data and assign values to it; the data is then processed by hidden neuron layers responsible for decision-making. The more data you feed to a neural network, the smarter it gets. A voice capturing device is the key component of voice-controlled Home Automation solutions

Should You Develop Custom Speech Recognition Software or Use a Ready-made Solution?

There’s a plethora of speech recognition programs available on the market, including Siri, Alexa, Google Assistant and Cortana. With the exception of Microsoft’s Cortana, all assistants boast a solid market presence and a relatively easy set-up; Google Assistant and Alexa have made their SDKs public and now work in sync with a wide range of 3rd-party devices including laptops, smartphones, wearables and home appliances. When it comes to performance, all assistants (again, excluding Cortana) do fine in basic voice recognition, although they might fail to distinguish between homonyms and interpret commands in noisy environments.
Have any questions? Ask our team!
Having said that, the speech recognition programs have their strengths and limitations. Alexa has developed a larger number of 3rd-party skills and can easily book a ride with Uber and order a Pizza from Domino’s; however, it’s too reliant on specific terminology, which means it simply does pattern-matching rather than make data-driven decisions. Google Assistant and Siri are by far the most intelligent assistants on the market; yet, they are seamlessly merged into their proprietary systems (Android Things and HomeKit), which could be considered as a disadvantage due to interoperability reasons.
Also, there are open-source ready-made voice assistant solutions like Mycroft, CMU Sphinx and Jasper with excellent documentation and small, but growing dedicated developer communities – and even more speech recognition libraries and APIs including IBM Speech to Text,, Google Speech Recognition and Python’s SpeechRecognition package which allow Home Automation start-ups to create custom voice recognition solutions. Since the major goal of a voice-controlled connected home product is to be able to understand user commands and provide context-based follow-ups, it would take a long time and resources to train your custom speech recognition solution; unlike Amazon and Google, you don’t have unlimited financial resources, continuous access to customer data and 3 to 12 months of development to improve the accuracy of your program).

4 Things to Consider before Building a Voice-controlled Home Automation System

Most 3rd-party AI Assistants Only Perform Well under Ideal Circumstances

Recognizing human speech is really hard. Companies eyeing the promising market have to overcome some major challenges including microphone-related issues, sound overlapping, background noise and pronunciation variations. As of now, even well-trained personal assistants like Google, Siri and Amazon only understand voice commands when the room is quite – and there are two ways to solve this problem:
  • Legacy approach: you can enhance your voice-controlled connected home solution with a noise reduction chip – for instance, Qualcomm’s Snapdragon 845 – during the hardware design stage. This strategy, however, can only be applied to removing constant noises from fixed locations. To be able to control a Home Automation system from any place in a house, your customers perhaps need a more flexible solution.
  • Smart approach: once again, neural networks come to rescue. A speech recognition program can be trained to identify the physical attributes of a human voice and ignore everything else. That’s exactly what Cypher, a US start-up whose custom AI algorithm improved Alexa’s performance in noisy environments by over 120%, did. However, the accuracy of an AI-powered speaker is still affected by its ability (or lack thereof!) to determine the exact position of a sound source.

The majority of voice-based Home Automation products are seamlessly merged into their proprietary systems

AI’s Cognitive Capabilities are Still Limited

Even though the accuracy of the voice recognition technology is now estimated at 95% and continues to improve, even the best home assistants on the market can only return simple answers to simple searchable questions about the weather, currency rates and movies playing in local theatres. This should be fine if you only want to enable routine Home Automation that doesn’t go beyond switching the lights and turning up connected thermostats via voice commands.
Major problems occur, however, if you intend to use AI for security system management – for instance, ask your voice assistant who’s at the door right now (meaning you need a camera and a well-trained image recognition algorithm capable of identifying familiar faces) or turn to biometrics and create a voice-based ID authentication solution.
You should also consider the tasks your potential customers might want to complete with the help of a Smart Home system – for example, manage emails and personal calendars, make purchases (62% of users who own voice assistant devices are actually willing to do so) and interact with 3rd-party gadgets and mobile applications on demand.

Customers Don’t Need Another Standalone Smart Home Gadget

As we mentioned earlier, over 47 million US consumers have access to a smart speaker; 19.3% of those own two devices. 17.5 million US homes – or 13% of all broadband households in the country – have a smart thermostat, too. By 2020, almost 60 million of US households will have an AI-powered speaker However, most IoT devices only talk to each other through middleware as they use different communication protocols and interaction patterns by design. Examples of such gateway solutions include the If This Then That (IFTTT) applet service, MIT’s Oxygen and Amigo Ambient Intelligence. As a forward-thinking entrepreneur, you shouldn’t lure your potential customers into another single-vendor environment; instead, you should provide open-source API and enable support for as many connected gadgets as possible.

Make Sure You Understand Home Automation Development Costs Upfront

Home Automation solutions rely on cloud-based infrastructure and apps supporting sensor data processing and visualization. There’s also hardware which normally amounts to 35-80% of the total Smart Home project costs. And don’t forget about not-so-obvious infrastructure expenses like software licensing and cellular network providers’ fees. At the end of the day, you cannot build a prototype of an IoT solution for less than $ 50 thousand (the estimate is based on the median Eastern European developer hourly rates ranging from $ 30 to $ 35).
The cost of building an IoT solution may be high, but it doesn’t change the fact that voice-controlled Home Automation is a very lucrative market segment. By 2020, 50% of all searches made online will be voiced-based. Besides routine Home Automation, voice-controlled Smart Home systems can be potentially used in disabled and elderly people care. Thanks to advancements in NLP and speech recognition, you don’t even have to build AI-based software from scratch!
All that’s left is to address a reliable IoT development company, right?
Did you like our post?

Subscribe to our monthly IoT digest!

Please, enter a valid email address
Please, agree with our privacy policy

Check out services we provide!

Get a Free Quote Now!

Popular Posts

Link copied to clipboard

Get 3 bonuses now!

Just Share Your Project Idea
(We'll Keep It Safe!)
& Get Your Bonuses for Free!

get for free now