Note: Your custom schedule will not be saved unless you create a new account or login to an existing account.
| Paper: | SP-L6.4 |
| Session: | Feature Analysis for Speech Recognition |
| Time: | Thursday, May 20, 14:00 - 14:20 |
| Presentation: |
Lecture |
| Topic: |
Speech Processing: Feature Extraction |
| Title: |
ROBUST SPEECH FEATURE EXTRACTION BY GROWTH TRANSFORMATION IN REPRODUCING KERNEL HILBERT SPACE |
| Authors: |
Shantanu Chakrabartty; Johns Hopkins University | | |
| | Yunbin Deng; Johns Hopkins University | | |
| | Gert Cauwenberghs; Johns Hopkins University | | |
| Abstract: |
A robust speech feature extraction procedure, by kernel regressionnonlinear predictive coding, is presented. Features maximallyinsensitive to additive noise are obtained by growth transformation ofregression functions spanning a Reproducing Kernel Hilbert Space(RKHS).Experiments on TI-DIGIT demonstrate consistent robustness of thenew features to noise of varying statistics, yielding significantimprovements in digit recognition accuracy over identical modelstrained using Mel-scale cepstral features and evaluated at noiselevels between 0 and 30dB SNR. |
| |
| Back | |