Speakerfilter-Pro: an improved target speaker extractor combines the time domain and frequency domain.
Shulin HeHao LiXueliang ZhangPublished in: ISCSLP (2022)
Keyphrases
- frequency domain
- spatial domain
- fourier transform
- cross correlation
- feature extraction
- power spectrum
- spectral domain
- denoising
- subband
- facial asymmetry
- fourier domain
- speech recognition
- frequency domain analysis
- frequency analysis
- spectrum analysis
- power spectral
- power spectra
- frequency spectrum
- fourier analysis
- phase correlation
- low pass
- discrete fourier transform
- bandpass
- multiscale
- image processing
- automatic speech recognition
- target detection
- image quality