This is a technology to remove the noise regression effect in a system where speaker and microphone coexist. The noise sources are adaptively removed so as to enable voice recognition even when the sound sources are played, and thus, only the voice entered into microphone is extracted.
This variable learning rate noise removal algorithm based on NLMS (normalized least mean square) maintains the voice entered into the microphone in a natural state, while at the same time removing noise sources coming from the speaker. It is used for voice recognition pre-processing in media devices, such as smartphones, TVs, MP3 players and navigation systems.
- ANC based on least-mean square (LMS)
- ANC based on normalized LMS (NLMS)