Research Article

Segmentation and Classification of Vowel Phonemes of Assamese Speech Using a Hybrid Neural Framework

Algorithm 3

PNN and F1 based recognition algorithm.
(1) Input: Speech of size , sampling frequency , duration second
(2) Preprocess the signal using preprocessing algorithms described in Section 7.1
(3) Obtain Dw1, Dw2, Dw3, Dw4, Dw5, Dw6 using the DWT
   based segmentation described in Section 4 or obtain SW1, SW2, SW3, SW4, SW5, SW6 using the SOM weight
   vector extraction algorithm described in Section 5
(4) Find first formant frequency of Dw1, Dw2, Dw3, Dw4,
Dw5, Dw6 and store as , , , , and or Find first formant frequency of SW1, SW2, SW3, SW4,
SW5, SW6 and store as , , , , and
(5) Load PNN1
(6) Decide VOWEL-A
If Dw1 = VOWEL-A and = F1 of vowel
else if
Dw2 = VOWEL-A and = F1 of vowel
else if
Dw3 = VOWEL-A and = F1 of vowel
else if
Dw4 = VOWEL-A and = F1 of vowel
else if
Dw5 = VOWEL-A and = F1 of vowel
else if
Dw6 = VOWEL-A and F1 of vowel
else Decide
“Not Assamese vowel Phoneme /i/”.
or
Decide VOWEL-A
If SW1 = VOWEL-A and F1 of vowel
else if
SW2 = VOWEL-A and F1 of vowel
else if
SW3 = VOWEL-A and = F1 of vowel
else if
SW4 = VOWEL-A and = F1 of vowel
else if
SW5 = VOWEL-A and = F1 of vowel
else if
SW6 = VOWEL-A and = F1 of vowel
else Decide
“Not Assamese vowel Phoneme /i/”.