Towards pre-processing for improving speech recognition efficiency High-quality Audio Middleware

Grape Systems Inc.

This is a demo for speech recognition, triggered when music is playing in the environment.

The M3 Board’s speech recognition engine will recognize the trigger word “Alexa” when it is spoken while music is playing with 4 microphone modules connected to the R-Car M3 board.

When the trigger word is correctly recognized, the speaker will release a “ping” sound and the music’s volume will lower for approximately 4 seconds.
Afterwards, the music’s volume will return, and the trigger word will be recognizable again.

We offer the software library for Renesas processors, including noise removal processing which does not pick up music and surrounding noise before speech input to the speech recognition engine, and processing picking up speech from remote locations.