Vowel, Digit and Continuous Speech Recognition Based on Statistical, Neural and Hybrid Modelling by Using ASRS_RL

Dumitru, Corneliu Octavian; Gavăt, Inge

doi:10.1109/eurcon.2007.4400336

Cited by 6 publications

(4 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In Table 1, are displayed only the results in the case of training MS and testing with MS and FS. Similarly results were obtained for the training with FS [2].…”

Section: Vowel Recognitionsupporting

confidence: 75%

NN and Hybrid Strategies for Speech Recognition in Romanian Language

Dumitru¹,

Gavăt²

2008

Proceedings of the 4th International Workshop on Artificial Neural Networks and Intelligent Information Processing

View full text Add to dashboard Cite

In this paper we present results obtained with learning structures more "human likely" than the very effective and widely used hidden Markov model. Good results were obtained with simple artificial neural networks like the multilayer perceptron or the Kohonen maps. Hybrid structures have proven also their efficiency, the neuro-statistical hybrid applied enhancing the digit recognition rate of the initial HMM. Also fuzzy variants of the MLP and HMM gave good results in the tested tasks of vowel recognition.

show abstract

“…In Table 1, are displayed only the results in the case of training MS and testing with MS and FS. Similarly results were obtained for the training with FS [2].…”

Section: Vowel Recognitionsupporting

confidence: 75%

NN and Hybrid Strategies for Speech Recognition in Romanian Language

Dumitru¹,

Gavăt²

2008

Proceedings of the 4th International Workshop on Artificial Neural Networks and Intelligent Information Processing

View full text Add to dashboard Cite

show abstract

“…Each hybrid model is made of 5 states, each state being associated with one output node of the MLP. The MLP has one hidden layer (100 nodes), and the input layer consisting of 12 nodes (Dumitru et al, 2007). The results for digit recognition are presented in Table 13.…”

Section: Telephone Dial Applicationmentioning

confidence: 99%

“…Continuing this tradition, our team has experimented a set of Computational Intelligence algorithms for Automatic Speech Recognition (ASR) extended also for speech understanding in the frame of the three main paradigms mentioned in the introduction. (Dumitru et al, 2007). For DDEL the database used is the AMP database (www.amp.ece.…”

Section: Techniques Applied For Romanian Languagementioning

confidence: 99%

Progress in Speech Recognition for Romanian Language

Dumitru¹,

Gavăt²

2008

Advances in Robotics, Automation and Control

Self Cite

View full text Add to dashboard Cite

“…For continuous speech recognition there are two databases namely CDRL (Continuous Database for Romanian Language) and SCDRL (Second Continuous Database for Romanian Language). The characteristics (Dumitru, Gavat, 2007) for the first database CDRL are the following: the database is constituted for training by 3300 phrases, spoken by 11 speakers, 7 males and 4 female speakers, each speaker reading 300 phrases, and for testing by 880 phrases spoken by the same speakers, each of them reading 80 phrases. The training database contains over 3200 distinct words, while the testing database contains 1500 distinct words.…”

Section: Gender Trained Modelsmentioning

confidence: 99%