Publication Details
Phoneme Recognition using Temporal Patterns
Schwarz Petr, Ing., Ph.D. (DCGM)
Heřmanský Hynek, prof. Ing., Dr. Eng. (DCGM)
Černocký Jan, prof. Dr. Ing. (DCGM)
speech recognition, feature extraction, temporal patterns
We investigate and compare several techniques for automatic recognitionof unconstrained context-independent phoneme strings from TIMIT andNTIMIT databases. Among the compared techniques, the technique based onTempoRAl Patterns (TRAP) achieves the best results in the clean speech,it achieves about 10% relative improovements against baseline system.Its advantage is also observed in the presence of mismatch betweentraining and testing conditions. Issues such as the optimal length oftemporal patterns in the TRAP technique and the effectiveness of meanand variance normalization of the patterns and the multi-band input theTRAP estimations, are also explored.
@inproceedings{BUT14184,
author="Pavel {Matějka} and Petr {Schwarz} and Hynek {Heřmanský} and Jan {Černocký}",
title="Phoneme Recognition using Temporal Patterns",
booktitle="Proc. 6th International Conference Text, Speech and Dialogue, TSD2003",
year="2003",
pages="465--472",
publisher="Springer Verlag",
address="Ceske Budejovice",
isbn="3-540-20024-X"
}