Speech enhancement method and device applicable to strong noise environments

The invention discloses a speech enhancement method and device applicable to strong noise environments. The method comprises establishing an air conduction speech sensor speech detection and non-air conduction speech sensor speech detection combined model under a clean environment; estimating a noise model of a currently received air conduction speech sensor speech detection signal according to a speech signal detected by a non-air conduction speech sensor; correcting parameters of the combined model by using the obtained noise model; and enhancing the air conduction speech sensor speech detection signal through the combined model before and after correction, and outputting enhanced speech signals. The device comprises an air conduction speech sensor, the non-air conduction speech sensor, a multi-channel data acquisition module, a noise model estimation module, a combined model correction module, a speech enhancing module and a training and self-adaptive module for the combined model. Compared with the prior art, the speech enhancement method and device have the advantages of being strong in noise immunity, good in speech quality and the like.