How many times have we watched an action movie and been forced to resort to subtitles due to inexplicably mismatched volumes of dialogue, music, and sound effects? A recent CBS News poll revealed that over half of Americans use closed captions at least some of the time. This suggests that many people struggle to understand dialogue at their preferred listening volumes. Skyworks has created a solution that addresses this recurring problem without degrading the quality of non-speech sounds: Skyworks AI Dialog Boost.
Based on Skyworks’ AI Noise Reduction model, this innovative solution separates narration, dialog, and other forms of speech from background music and sound effects then subsequently remixes in real-time. Opposed to traditional algorithmic approaches of boosting dialogue, which often blindly alter frequency and dynamic range, degrading the quality of non-speech sounds, Skyworks’ AI Dialog Boost uses speech separation and enhancement, without adding any latency when integrated into existing audio processing paths.
The advanced AI that isolates speech from a general audio signal is based on a proprietary award-winning neural network that is trained on an expansive variety of speech and noise signals. The solution is scalable from enterprise to edge, and in a typical implementation it meets the needs of the market by requiring approximately 3M parameters and adding no latency when integrated into an existing audio processing path.
Designed as a scalable algorithm, Skyworks AI Dialog Boost can be tailored to specific embedded applications. The solution runs on a Tensilica HiFi4 core at 600MHz, requiring less than 80% CPU load, and operating at 48kHz in stereo. And since this process happens in real-time, system designers can allow the end user to set the level of the remix according to their preferences.
The integration of Skyworks AI Dialog Boost in microphones or video devices can enable in-camera processing that effectively minimizes unwanted sounds like wind, handling, and environmental noises, such that a camera system can capture high-quality speech on the fly with no post processing.
“AI is a game changer in this application, offering the unique ability to pull the voice out of an existing mix in real time, which was something only previously dreamed about. This will make a huge impact for an array of applications such as TVs and soundbars where it will provide system designers the ability to enable custom adjustments and provide improvement for people with hearing loss, while also benefiting individuals without hearing loss,” says Shawn Scarlett, Marketing Director for audio and the AI systems group at Skyworks.
“Skyworks is at the forefront of using AI to innovate beyond classic signal processing techniques, and our AI Dialog Boost represents a fitting example of this evolution. Delivering break-through performance while maintaining scalability for embedded applications, it will also complement our new SM4G Soundbar Processor making it easier to bring to market a high quality and feature-rich soundbar.”
Skyworks AI Dialog Boost will be a critical piece of the upcoming SM4G Soundbar Processor, bolstering Skyworks’ industry recognition in the consumer audio segment.
www.skyworksinc.com
