Case Study: Real-time Noise Suppression on Edge Devices Using a Practical AI-based Approach
We’re thrilled to be featured on Arm’s new case study regarding #Ignitarium’s Noise Suppression IP for Arm Cortex-M processors.
Ignitarium has come up with a deep learning-based real-time noise suppression software that can run on low-cost microcontrollers. To counter stationary and non-stationary noises, a traditional DSP approach is not sufficient. This requires certain hybrid models including a deep learning-based approach. Using deep learning, it is possible to train a model with a variety of conditions and the resulting model can provide good results in these varying environments. Ignitarium’s Real-time Noise Suppression algorithm (IGN-RNS) is one such approach, where a linear regression method is used to denoise noisy speech input in real-time.
“Drawing on our strength in multimedia, DSP, and the many years we spent on video AI, it was natural for us to see what we can do with AI in the field of audio analytics. We understand semiconductors and the MCU/SoC ecosystem very well. This opened up an interesting opportunity to implement real-time audio analytics on embedded devices, paving the way for the first engagement with a lead customer,” said Sanjay Jayakumar, CEO of Ignitarium.