Digital Image Processing
Editorial note: Bridging the gap between human and automatic speech recognition
Speech Communication
Evaluation of Objective Quality Measures for Speech Enhancement
IEEE Transactions on Audio, Speech, and Language Processing
| Hi-index | 0.00 |
In this paper a method to remove noise in speech signals improving the quality from the perceptual point of view is presented. It combines spectral subtraction and two dimensional non-linear filtering techniques most usually employed for image processing. In particular, morphological operations like erosion and dilation are applied to a noisy speech spectrogram that has been previously enhanced by a conventional spectral subtraction procedure. Anisotropic structural elements on grayscale spectrograms have been found to provide a better perceptual quality than isotropic ones and reveal themselves as more appropriate for retaining the speech structure while removing background noise. Our procedure has been evaluated by using a number of perceptual quality estimation measures for several Signal-to-Noise Ratios on the Aurora database.