Amazon's Alexa Voice Service Integration for AWS IoT Drops the Hardware Requirements Dramatically
Alexa voice services can now run on devices with as little as 1MB of RAM and a Cortex-M processor — with a little cloud help.
Amazon has announced that its Alexa voice-activated assistant service is coming to a wider range of low-powered devices, thanks to a version dubbed Alexa Voice Service (AVS) Integration for Amazon Web Services (AWS) Internet of Things (IoT) Core which can run on gadgets with as little as 1MB of RAM.
While Amazon's Alexa ecosystem — which allows gadgets to respond to a trigger phrase, recognise speech, and respond in kind or by activating compatible peripheral devices like smart door locks and thermostats — is popular among gadget builders, it's always needed some fairly hefty hardware: Anyone building an Alexa-compatible device needed a platform with a minimum of 100MB of free RAM and an Arm Cortex-A processor. Now, AVS Integration for AWS IoT Core — as the extended platform is somewhat clunkily known - reduces those requirements down to just 1MB and an Arm Cortex-M microcontroller.
In a press release seen by VentureBeat but, oddly, not yet released publicly, Amazon claims the new platform requirements reduce the materials cost by up to 50 percent - meaning that it may be possible for Alexa to extend beyond smart speakers and other relatively highly-priced devices into a world where even a lightbulb comes with its own microphone, speaker, and Alexa support.
The system works by offloading the heavy lifting of speech recognition to Amazon's Web Services (AWS) cloud platform via MQTT. It also stands separate to the Amazon Alexa Skill for Arduino IoT Cloud — the latter being aimed more at makers and hobbyists, the former being targeted at companies looking to build low-cost Alexa devices for retail sale.
At the same time, Amazon announced the extension of its AWS IoT Greengrass service with Docker container support and a Stream Manager feature which allows for the collection, processing, and exportation of data streams from edge devices without the need to write AWS Lambda functions as before.
Other new features include fleet provisioning functionality in AWS IoT Core and a secure tunnelling system which provides the ability to communicate with devices even when behind a restrictive firewall. "If [an] MRI machine in hospital or something in factory stops working," Amazon's Dirk Didascalou explains to SiliconAngle, "but it’s in a private network or behind a firewall, this allows access."
Amazon is expected to offer more details on the new features at AWS re:Invent, its annual conference, next week.