摘要 |
A system for Automatic Speech Recognition (ASR) to eg. wake up the applications processor AP 32 on a mobile phone 12 detects a first part of a trigger phrase (TP1, fig. 3) at a detector 40 and sends a control signal (Acvt) to activate a speech processing block 42 for eg. adapting parameters such as mic beamforming or noise reduction until convergence (Coeff, fig. 3). If a second portion TP2 of the trigger phrase is detected 38, the activation of the speech processing block is maintained, while a deactivation signal is sent if the second portion is absent. |