Enhancing neural network based consonant recognition using formant transitions in neighbouring vowels
Shyamala C. Sivakumar, William Robertson · 1992
This paper proposes an ANN architecture for consonant recognition in continuous speech that utilizes cues in the neighbouring vowels, in addition to cues in the consonants themselves. The additional cues are due to coarticulatory effects and enhance the recognition capability, especially for stop consonants which have weak speech cues. Simulation results using real speech data demonstrate the improvement in performance.>