The IoT has sparked a love affair with smart, connected things designed to
bring more convenience, ease, efficiency and safety to our lives. While we
appreciate the added benefits of our IoT-connected thermostats, doorbell
cameras and wearable devices, there is still great opportunity to make these
‘things’ smarter ... more intuitive, more tuned to how humans actually
communicate, more like our Alexa, Siri or Google Assistant-enabled smart
speakers.
Voice – human speech – is the most natural HMI. We want voice control, and
crave more of it, along with greater speed and accuracy. Why tediously type
the title of a Netflix movie on your smart TV interface when you can simply
talk to a handheld remote to find your favorite movie? Why get up from the
couch and walk across the room to dim the lights when you can ask your
voice-enabled light switch for help? While touch interfaces aren’t going away
anytime soon, we’re seeing more use cases for voice as an easier way to
control devices and systems in our homes, offices, cars and public spaces.
Even more so during these times of COVID-19.
According to an ABI Research, by 2025, approximately 14% of all smart home
products will be conversational AI enabled, and will grow more than 5x faster
than the overall smart home market. Key growth drivers include the growing
consumer demand for smart devices, the proliferation of voice-enabled devices
and the rise of voice-enabled automotive infotainment systems. However, the
cost and expertise of adding voice enablement to smart devices is a major
factor when designing these types of devices.
To bring affordable voice control to more and more devices, NXP Semiconductors
is committed to reducing the design effort for our customers to further help
reduce the cost and complexity of the overall system solutions. Our goal is to
help developers everywhere add voice control to almost everything, from
consumer electronics and home control applications to industrial and
automotive applications. The use cases are endless. The key is to offer a
comprehensive portfolio of silicon- and software-based voice solutions scaling
across NXP’s EdgeVerse™ processing continuum. We have added
Retune DSP, a leading provider of digital signal processing (DSP) software solutions
for voice control and communication, to our voice portfolio.
Headquartered in the greater Copenhagen area, Retune DSP provides highly
accurate, small-footprint, low-power voice processing solutions that are
popular with a growing customer base. We believe Retune offers the right DSP
software technology to help NXP reach the next level in delivering
high-performance voice control solutions to customers worldwide.
Retune DSP is Now Part of NXP
Renowned for the quality of their DSP software and algorithms, Retune DSP is
one of the best voice technology companies in today’s market. The company has
built a strong reputation and loyal customer base as a one-stop shop and
full-service software supplier delivering a complete voice control stack
including multi-microphone beamforming, multi-channel acoustic echo
cancellation and wake-word engines.
When developers required extremely high levels of detection accuracy within
tiny model sizes, Retune DSP stepped up and worked closely with customers to
deliver the right voice solution for their application needs. Retune DSP’s
three voice DSP software products will merge with NXP’s current voice
solutions portfolio. This includes:
Hands-Free Telephony Suite:
-
Conversa is a high-performance, multi-microphone, full-duplex voice
processing software suite that provides total control of uplink and downlink
audio. Conversa delivers exceptional sound quality for hands-free speech
processing applications including Bluetooth speakerphones, conference
phones, headphones and headsets, true wireless stereo (TWS) earsets and
automotive voice solutions.
Voice Control Audio Front End Suite:
-
VoiceSeeker is a state-of-the-art multi-microphone voice control audio front-end
DSP solution offering high-performance beamforming and multi-channel
acoustic echo cancellation to deliver exceptional trigger word detection and
payload accuracy. Retune’s ultra-efficient algorithm can simultaneously
process up to 24 independent acoustic echo cancellers per core, unlocking
the potential of voice control in multi-channel home theater environments.
Effective beyond 10 meters, VoiceSeeker’s beamformer technology is ideal for
far-field voice control applications.
Wake-Word Engine:
-
VoiceSpot is a highly accurate wake-word engine leveraging Retune’s
proprietary ML-based pattern recognition technology. Retune’s
high-performance voice triggers have ultra-small memory footprints,
requiring fewer MIPS than competing wake-word engines. The combination of
VoiceSpot and VoiceSeeker software provides a complete front-end voice
control solution, enabling developers to move quickly through the software
integration process. VoiceSpot also supports customer-defined voice trigger
words or phrases.
In Tune with NXP’s Edge Processing Continuum
Retune’s DSP software is an ideal fit for our processor continuum –
complementing the broad spectrum of voice control solutions offered by NXP and
our voice ecosystem partners. We believe this strategic acquisition will
complement the NXP brewed voice intelligent technology (VIT) wake-word engine
with its audio front end, bringing scale and new market opportunities to
Retune’s market-proven voice processing, beamforming, acoustic echo
cancellation and wake-word engine technologies.
NXP offers a broad range of solutions for adding voice control directly to any
device. From flexible implementations to turnkey solutions, customers can
choose the processing platform that meets their application needs and BOM cost
target and can easily move up the edge processing continuum to larger memory
options, more powerful Arm® cores, integrated DSPs and
additional on-chip accelerator and security options.
To accelerate time to market, we provide comprehensive development
environments including our popular MCUXpresso SDK and fully functional example
applications enabling fast, easy evaluation. We also offer powerful ML
technologies for customers developing in-house voice applications. Additional
specialized solutions are available from NXP’s leading front-end and voice
recognition partners.
With the combination of Retune’s high-performance DSP software and NXP’s broad
customer base, global footprint and scale, we are well-positioned to deliver
breakthrough voice solutions to more customers through our combined R&D, sales
force and customer support teams.
“Hey NXP, what’s next?”
Discover the full range of NXP’s comprehensive voice solutions at
nxp.com/voice
