Classification-based close talk speech enhancement

This paper addresses the problem of close talk speech enhancement as a binary classification using dual microphones features in noisy and reverberant environments. In this work, we investigate a speech segregation framework, in which deep neural networks (DNN) are employed as a mechanism to find the robustness classifier from two microphones inputs. The paper reports the successful attempt to use dual microphones signals and energy difference features and monaural features as the segregation cues with this framework. Results with recording corpus show that robust performance can be achieved across a variety of multi location, noise types and reverberant conditions.