Exponentiated magnitude spectrogram-based relative-to-maximum masking for speech enhancement in adverse environments.
Chen-Li LinZi-Qiang LinSyu-Siang WangYu TsaoJeih-Weih HungPublished in: ICCE-TW (2020)
Keyphrases
- speech enhancement
- speech signal
- single channel
- noisy environments
- noise reduction
- multi channel
- speech recognition
- signal to noise ratio
- linear prediction
- spectral analysis
- dynamic environments
- frequency domain
- wiener filter
- background noise
- automatic speech recognition
- image segmentation
- multiresolution
- wavelet decomposition
- edge preserving
- information retrieval
- image acquisition
- prior information