Using cloud-based systems to add voice recognition is an option but can present user privacy and latency issues. They also require end users to have accessible networks and have a tendency to be cost-prohibitive for many smart home and consumer IoT applications.
To help designers short circuit some of these challenges, NXP is bringing offline voice control to the edge.
The NXP EdgeReady SLN-LOCAL-IOT solution is based on the i.MX RT family of crossover MCUs and comes with everything a developer needs for out-of-the-box concept development. Its ultra-small form-factor and turnkey nature make it an ideal platform for adding voice to smart products with an eye toward minimizing risk, time-to-market, and development effort.
The system brings a high quality far-field audio front end, customizable wake word engine, and command recognition engine to the i.MX RT. Together with the full audio hardware design, this new solution works with application-specific speech recognition models.
Offline implementation means eliminating some of the elements that drive up costs, like Wi-Fi modules and cloud service fees. Running instead on NXP’s low-cost i.MX RT106L MCU, this system delivers a one-two punch of all-in-one functionality and cost accessibility.
The SLN-LOCAL-IOT solution incorporates many cutting-edge technologies typically found in much more expensive hardware and coprocessor DSPs. Leveraging the performance of i.MX RT microcontrollers, this solution can accomplish most, and in many cases all, of the audio capabilities typically offered in MPU+DSP designs.
Another key benefit is its privacy-by-design, which means none of the audio gets transmitted to the cloud – all processing is done locally on the device itself, no voice recording are stored permanently, anywhere.
The audio processing front end and the local control library are the unique enabling technologies and these two features can be used together or separately to customize the user experience. The local control library package features both wake word and command detection and is easy to integrate into any application. Plus, it uses less than 100KB of RAM for typical models, leaving plenty of RAM for the rest of the application.
After setting up and initializing the library, the application simply feeds an input audio stream into the control library. As the library detects the wake word or command, it executes callbacks for the user application to handle them.
Feeding the library is the audio processing front-end. This component is responsible for listening to multiple microphones (up to three on the voice solution) and cleaning up the audio through beam forming and echo cancellation. The front-end chooses the best beam and sends the audio to the library.
By bringing voice control to the edge, you can introduce features consumers have come to expect without sacrificing performance, cost, or privacy.
|Hands-free voice control without cloud connectivity|
|– Smart lighting
– Shade and fan controls
– Smart plugs and outlets
– Smart appliances
– Set-top boxes and residential gateways
|– Alarms and thermostats
– Garage door openers
– Air conditioners
– Point-of-sale terminals
– Industrial Automation
– Hands-free process control
Soon it will be possible to combine the SLN-LOCAL-IOT with other leading AI/ML capabilities including face recognition, object detection, and anomaly detection to enable a variety of exciting new applications.
For additional information, and to get your hands on the SLN-LOCAL-IOT development the kit and i.MX RT106L MCU, see nxp.com/mcu-local.
Just as its name implies, an NXP EdgeReady solution is a complete, ready-for-the-edge platform. Each NXP EdgeReady solution is production-grade and designed to deliver quick and easy product development. NXP EdgeReady solutions include hardware optimized for cost and form factor in addition to fully tested and integrated self-contained NXP-certified software. Source code, schematics, and layouts are included, as well as onboard machine learning software that can be modified, tuned, and trained for unique applications. See more EdgeReady solutions.