Voice Isolation Technology: A Feature That Speaks for Itself

Tuesday 5/10/22 08:31am
|
Posted By Rami Husseini
  • Up0
  • Down0

Qualcomm products mentioned within this post are offered by
Qualcomm Technologies, Inc. and/or its subsidiaries.

Hands up if you were on a video call this week? And keep your hands up if you got distracted by someone on the call typing, their dog barking, their kids playing, or other background noise. You were not alone.

These experiences have become normal in our working environments over the last few years and are a big reason why our Windows ACPC demos at Mobile World Congress (MWC) 2022 were a big hit. We showcased our latest AI-powered noise suppression on the new Lenovo ThinkPad X13s, which can effectively remove background noise while someone speaks.

This video demonstrates how well the technology can isolate a speaker’s voice from different background noises ranging from crinkling snack bags to drums.

These algorithms, found in the Qualcomm Audio & Voice Communication Suite, bring great audio and voice enhancement advances. Our ML algorithms isolate the user’s voice, rather than suppressing background noises, without needing multiple microphones as would normally be necessary using traditional techniques.

They’re supported by the Snapdragon 8cx Gen 3 Mobile Compute Platform, our latest compute platform for premium Windows devices, powered by Snapdragon. It’s just one of the Windows on Snapdragon solutions you can use for Windows development.

Behind the scenes, neural networks trained to identify background noises are hard at work analyzing the incoming audio stream in real-time. They’re powered by the Qualcomm AI Engine, and more specifically, the Qualcomm Hexagon DSP (our optimized neural processing unit or NPU), to perform voice isolation with the utmost power efficiency. The heterogeneous design allows for rich concurrency control and frees the SoC’s Qualcomm Kryo CPU and Qualcomm Adreno GPU to handle other processing tasks like working on your presentations. Most importantly, this architecture allows for several AI models to run concurrently with high levels of performance per watt, essential for long battery life and acceptable thermal dynamics on today’s mobile devices. Examples of AI models that can run concurrently include background blur or gaze correction, as well as auto framing to help keep your face on camera. Check out this great video to see how auto framing can work.

Clear audio is crucial due to today’s reliance on video conferencing and remote meetings. But it doesn’t just stop there. There are all kinds of possibilities ranging from clear inter-player communication for multi-player games to easier talk to text or voice assistant use in noisier locations.

Development
For hardware developers looking to build mobile Windows devices and software developers building apps for them, this functionality is available from the Snapdragon 8cx Gen 3. Microsoft recently announced that Windows 11 will natively support effects like noise eye correction and noise suppression in video calls on devices with dedicated NPUs (e.g., the Hexagon on the Snapdragon 8cx Gen3). And stay tuned for Microsoft Build 2022 where some new and exciting news around Windows on Snapdragon is currently planned! We have two great sessions scheduled for the event that we encourage you to check out.

Developers interested in Windows on Snapdragon development can also check out the Snapdragon Developer Kit (aka ECS LIVA Mini Box QC710 Desktop) to help port apps to the platform. This cost effective, small-form-factor development kit (available for purchase from Microsoft) allows you to test and debug applications remotely. This can be useful while waiting for your Snapdragon powered Windows device to be finalized during development or supplementing your existing development hardware.

Development for Snapdragon compute platforms is complemented by a suite of tools – both open source and from Microsoft – including LLVM, Sysinternals for ARM64, and of course, developers’ much-loved Visual Studio IDE. You can use Visual Studio 2022 and Visual Studio Code to build Windows 11 applications that run on devices powered by our compute platforms.

Conclusion
With society’s increasing dependence on staying connected remotely and the potential for new voice isolation applications, clear audio sits front and center for quality user experiences. The Snapdragon 8cx Gen 3, with our Qualcomm cVc Echo Cancelling and Noise Suppression (ECNS) technology, is poised to deliver these capabilities on premium Windows on Snapdragon mobile devices like laptops and tablets. Be sure to check out the various links in this blog so you too can hear the result that speaks for itself!

For additional information about Windows on Snapdragon development, be sure to check out the following developer-focused videos from ARM DevSummit 2021:

Also be sure to check out the following resources:


Snapdragon, Qualcomm Audio & Voice Communication Suite, Qualcomm Hexagon, Qualcomm Adreno, Qualcomm AI Engine, Qualcomm Kryo, and Qualcomm cVc are products of Qualcomm Technologies, Inc. and/or its subsidiaries.