Comparison of neural networks and conventional techniques for automatic recognition of a multilingual speech database
| Autor: | Juan Ríos, Darío Maravall Gómez-Allende, A. Carpintero, M. Pérez-Castellanos, J. Gómez-Calcerrada |
|---|---|
| Rok vydání: | 2006 |
| Předmět: |
Vocabulary
Database Artificial neural network business.industry Time delay neural network Computer science Speech recognition media_common.quotation_subject Deep learning computer.software_genre Backpropagation Recurrent neural network Multilayer perceptron Artificial intelligence Types of artificial neural networks business computer media_common |
| Zdroj: | Lecture Notes in Computer Science ISBN: 3540545379 IWANN |
| DOI: | 10.1007/bfb0035917 |
| Popis: | The paper presents an exhaustive test between the Multilayer Perceptron Fully Connected (MLPFC) trained with backpropagation using the feedback learning rule on one side and three well-established non-NN, standard Isolated Word Recognition (IWR) techniques on the other, using to this aim a multilingual speech database formed by a common vocabulary in three languages: German, Italian and Spanish. This comparative testing has been centered on speaker-independent conditions and two different training subsets have been used. An additional multi-speaker test has been performed on the Spanish vocabulary in order to compare the performances of the MLPFC and backpropagation with feedback learning algorithm with the non-NN methods. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|