Blind Adaptive Mask to Improve Intelligibility of Non-Stationary Noisy Speech

This letter proposes a novel blind acoustic mask (BAM) designed to adaptively detect noise components and preserve target speech segments in time-domain. A robust standard deviation estimator is applied the non-stationary noisy identify masking elements. The main contribution of proposed solution use this statistics derive an adaptive information define select samples with lower proportion. Thus, preserving intelligibility. Additionally, no signals previously required non-ideal mask. BAM three competitive methods, Ideal Binary Mask (IBM), Target (TBM), Non-stationary Noise Estimation for Speech Enhancement (NNESE), are evaluated considering corrupted by noises six values signal-to-noise ratio (SNR). Results demonstrate that technique achieves intelligibility gains comparable ideal masks while maintaining good quality.