Authors:
Fréjus A. A. Laleye
1
;
Eugène C. Ezin
2
and
Cina Motamed
3
Affiliations:
1
Université du Littoral Côte d’Opale and Université d’Abomey-Calavi, France
;
2
Université d’Abomey-Calavi, Benin
;
3
Université du Littoral Côte d’Opale, France
Keyword(s):
Formant Analysis, Fuzzy Logic, Deep Belief Networks, Phoneme Recognition, Continuous Speech Segmentation, Fongbe Language.
Related
Ontology
Subjects/Areas/Topics:
Adaptive Signal Processing and Control
;
Artificial Intelligence
;
Computational Intelligence
;
Fuzzy Control
;
Fuzzy Systems
;
Hybrid Learning Systems
;
Informatics in Control, Automation and Robotics
;
Intelligent Control Systems and Optimization
;
Machine Learning in Control Applications
;
Neural Networks Based Control Systems
;
Signal Processing, Sensors, Systems Modeling and Control
;
Soft Computing
Abstract:
This paper reports our efforts toward an automatic phoneme recognition for an under-resourced language,
Fongbe. We propose a complete recipe of algorithms from speech segmentation to phoneme recognition in
a continuous speech signal. We investigated a strictly fuzzy approach for simultaneous speech segmentation
and phoneme classification. The implemented automatic phoneme recognition system integrates an acoustic
analysis based on calculation of the formants for vowel phonemes and calculation of pitch and intensity of
consonant phonemes. Vowel and consonant phonemes are obtained at classification. Experiments were performed
on Fongbe language (an African tonal language spoken especially in Benin, Togo and Nigeria) and
results of phoneme error rate are reported.