 |
ITRW on Non-Linear Speech Processing (NOLISP 05)Barcelona, Spain |
 |
|
ITRW on Non-Linear Speech Processing (NOLISP 05)Barcelona, Spain |
|
In this paper we propose an improvement in the genera- tion of the cryptographic-speech-key by using an heuristic consisting on the selection of the dimensions with the best performance for each of the phonemes. This selection can be made thanks that we know the phonemes of the spoken user passphrase. First, the mel frequency cepstral coefficients, (first and second derivatives) of the speech signal are calcu- lated. Then, an Automatic Speech Recogniser, which models are previ- ously trained, is used to detect the phoneme limits in the speech utter- ance. Afterwards, the feature vectors are built using both the phoneme- speech models and the information obtained from the phoneme segmen- tation. Next, the Support Vector Machines classifier, relying on an RBF kernel, computes the cryptographic key. By applying the phoneme-space- representation heuristic our results show an improvement of 24.26%, 18.85%, 16.56% for 10, 20 and 30 speakers, from the YOHO database, respectively, compared with our previous results.
Bibliographic reference. García-Perera, L. Paola / Nolazco-Flores, Juan A. / Mex-Perera, Carlos (2005): "A phoneme-space-representation heuristic to improve the performance in a cryptographic-speech-key generation task", In NOLISP-2005, 114-121.