Posted On: Nov 24, 2021

The Amazon Chime SDK lets developers add real-time audio, video, screen-sharing, and messaging capabilities to their web or mobile applications. The Amazon Chime SDK now offers machine learning (ML) based echo reduction to help improve audio experiences. Acoustic echoes disrupt meetings or conference calls when the sound played by the loudspeaker is picked up by the microphone and it circulates back into the call. The new ML-based echo reduction capability is designed to reduce acoustic echoes and preserve voice quality during double-talk conditions, when two or more people speak at the same time.

The Amazon Chime SDK echo reduction also includes Amazon Voice Focus, the technology developed to provide noise reduction in the Amazon Chime SDK. Amazon Voice Focus uses machine learning and models of speech and hearing to reduce background noises such as fans, lawnmowers, and barking dogs as well as foreground noises like typing and shuffling papers — so that noise doesn’t detract from conversations and engagements. Developers can configure their meetings with echo reduction capabilities via the CreateMeeting API from the Amazon Chime SDK. After configuration, developers additionally have to enable the feature at the client level by applying the appropriate ML model for echo reduction when attendees join the meeting.

The processing for Amazon Chime SDK echo reduction is performed in real time using WebAssembly (WASM) and single instruction multiple data (SIMD) for efficient operation on most modern computers and browsers. This offering is currently available in the Amazon Chime SDK for Javascript.

To learn more about the Amazon Chime SDK, please see the following resources: