Method and device for speech recognition in surroundings with varying noise levels

摘要

it is a process of voice recognition in a bruitu00e9 acoustic signal.it includes a stage (1) digital signal bruitu00e9 time frames, a frame (2) of temporal model containing the word, a stage (3) for pattern recognition is compared to reference parameters recorded in a space for reference, a stage (5) for research ru00e9itu00e9rative models of noise in the signal bruitu00e9 frame.a stage (50) of a transition between a new research model of noise and the former, and when a transition is detected, a step (33) to update the reference area, the phase (2) includes a step of adjusting model parameters (200, 210) of the new part model of noise.