1 / 8

Survey of ICASSP 2013 section: feature for robust automatic speech recognition

Survey of ICASSP 2013 section: feature for robust automatic speech recognition. Repoter : Yi-Ting Wang 2013/06/19. Robust automatic speech recognition. Over the years, much effort has been devoted on developing techniques for noise robust Automatic Speech Recognition(ASR).

gil
Download Presentation

Survey of ICASSP 2013 section: feature for robust automatic speech recognition

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Survey of ICASSP 2013section: feature for robust automatic speech recognition Repoter: Yi-Ting Wang 2013/06/19

  2. Robust automatic speech recognition • Over the years, much effort has been devoted on developing techniques for noise robust Automatic Speech Recognition(ASR). • The goal is to make the ASR system more resistant to the mismatch between training and testing condition. • Noise reduction techniques: • Speech enhancement at the signal level • Robust feature extraction • Adapting the back-end models.

  3. A robust frontend for ASR : combining denoising, noise masking and feature normalization

  4. Filtering on the temporal probability sequence in histogram equalization for robust speech recognition HEQ FHEQ

  5. Ideal ratio mask estimation using deep neural networks for robust speeh recognition

  6. Solve the problem in reverberant environments • It is well known that the distortions caused by reverberation, background noise, etc., are highly nonlinear in the cepstral domain. • Dereverberation via suppression and enhancement can be applied for reverberation compensation. • The drawback is undesirable if the late reverberation is not estimated precisely.

  7. Noise model transfer using affine transformation with application to large vocabulary reverberant speech recognition

  8. Joint sparse representation based cepstral-domain dereverberation for distant-talking speech recognition • Most existing sparse representation methods only consider sparse modeling in a single signal space, and few considers dictionary learning across different signal spaces.

More Related