The overall system first extracts embedded speaker identity features using a neural network model, then the deep neural network speech enhancement takes the augmented features as the input to generate the enhanced spectra. With the additional embedded features, the speech enhancement system can be guided to generate the optimal output corresponding to the speaker identity.

(1) Only two models are used to implement a speaker-aware SE system,thus effectively decrease the computation complexity. (2) When comparing with DDAE, SaDAE improves 7.868.17 qualityintelligibility scores in noise conditions, respectively. (3) The best qualityintelligibility improvements when comparing SaDAE with DDAE over testing speakers are 28.2721.00, respectively.

The denoising capability in assistive listening devices is limited by the volume,requires a new design to improve the speech enhancement function. The demand is one factor that leads to the development on robotsIoT, which normally provide a speech interface. The variated speakernoise environments degrade the quality of the provided services,remains an important issue for the industry.