Issue link: https://resources.mouser.com/i/1442826
5 AI ❝ ❞ far-field capable, which means that they can hear you across a room. This requires powerful digital signal processing (DSP) capabilities for noise suppression, echo cancellation, beamforming, and "barge-in" (the ability to be heard over music or speech playback). A few semiconductor companies sell DSP devices that combine far-field voice processing with a wake word engine, but most new designs now use general-purpose microprocessors (such as NXP's i.MX 8M series) with software implementations for far-field voice and wake word engine. While the wake word engine and far-field voice processing run on a powerful edge computing device, the NLU processing for services like Alexa is all done in the cloud. These cloud-based voice assistants can be extremely powerful, offering many tools to developers and users, to create skills and routines that implement multiple functions. The command, "Alexa, I'm home," could turn on lights, close a garage door, set a thermostat, close blinds, and start playing your favorite radio station. However, there are users and use cases that don't have, don't want, or don't need cloud connectivity for a wide variety of reasons, but still want voice control capability. Local speech processing can deliver voice control to devices and users without access to the cloud, either with fixed commands or, using more powerful MPUs, developers can offer NLU capabilities similar to a cloud experience. Until recently, voice control implementations required a dedicated DSP device and powerful multi-core MPU running Linux, but now NXP is the first to enable OEMs to embed Alexa or fixed local commands into their products using an MCU running an RTOS on the i.MX RT series of crossover MCUs, such as the new i.MX RT106A for Alexa built-in. These MCU solutions offer a low-cost alternative to traditional MPU based implementations like NXP's i.MX 8M family, which can deliver rich multimedia experiences, but require the use of Linux, which may be unfamiliar to many developers of many smart home products, from light switches to washing machines. Now developers have choices for their voice control implementations: If they need to give users a rich powerful voice assistant, they can leverage the capabilities of a cloud-based assistant like Alexa. Alternatively, if users need only simple hands-free capabilities without requiring cloud connectivity, OEMs can implement local commands in their products. Both types of voice assistants can be achieved with powerful MPUs enabling rich media experiences (e.g. multi-room music and video calling, or local NLU), or voice-only focused implementations can be built on low cost, RTOS based MCUs. ■ NXP MCU Solution Enables Alexa Everywhere By Rick Bye, Senior Marketing Manager, NXP IoT Solutions C loud-based voice assistants, led by Amazon's Alexa™, have become common in many homes, embedded in smart speakers like Amazon's Echo™ products. Soon, the newly emerging trend to build voice assistants directly into smart devices such as light switches, ceiling fans, appliances, smoke detectors, and thermostats, will make voice control ubiquitous throughout the home. Voice assistants conveniently and unobtrusively located throughout the home enable a family's entire network of connected smart devices to be controlled by voice commands from any room. For example, "Alexa, turn off all the lights," or even more powerful routines can be created such as, "Alexa, good morning," which can turn on music, start the coffee pot, adjust the thermostat settings, and more. Today, a few smart home device and appliance makers are introducing versions of their products with built-in voice assistants. Until now, the technology required to embed cloud voice assistants has required a powerful multicore microprocessor unit (MPU), similar to the application processor in smart phones, with large Flash and SDRAM memories, and complex power management. Such implementations have not been a fit for cost sensitive consumer devices, which has impeded the proliferation of smart home devices with built-in voice. NXP Semiconductors has previously announced the first MCU- based implementation of an Alexa client, based on a new member of NXP's popular i.MX RT crossover MCU family of devices. This new solution enables device makers to build Alexa into products using a low cost, low power microcontroller unit (MCU), a device that is typically already required in any connected smart home product, meaning that OEMs can now add voice to their products at very low incremental cost (not much more than the cost of the microphones). Running on Amazon FreeRTOS™, NXP's new MCU-based AVS solution leverages the power of AWS IoT Core to minimize the processing resources needed to build Alexa into a product. Compared to previous implementations running Linux with large memory footprints, requiring more than 50MB RAM and several Giga-Bytes of Flash, NXP's MCU solution needs less than 1MB Turnkey, low cost i.MX RT solution speeds time to market.