Example-based Cross-Modal Denoising

Here are the details and links of the attached videos

• Please hear the videos, while you watch them.
   
• Click on any thumbnail image below, to run each video.
   
• The noises types are numbered from 1 to 5 and they are:
    #1 - Music from the song Sweet Child of Mine by GNR. SNR = 0.07.
    #2 - Music from the song I Wish It Would Rain Down by Pill Collins. SNR = 0.09.
    #3 - Female speech. SNR = 1.05.
    #4 - Male speech, SNR = 2.4.
    #5 - White Gaussian noise. SNR = 1 (0dB).
   
• If you have trouble runing the videos, download the codec from http://www.divx.com/en/win

Speech (Digits): Noisy Input.

Speech (Digits): Cross-Modal Audio Visual Denoising (Our algorithm) Corresponding to the noisy sequences above.

Speech (Digits): Unimodal (Audio-only) denoising (Cohen and Bardugo's algorithm) Corresponding to the noisy sequences above.

Speech (Digits): Unimodal (Audio-only) denoising (example-based algorithm, run only on the audio channel) Corresponding to the noisy sequences above.