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

Algorithm 4

Preprocessing algorithm.

(1) Input: Speech 𝑆 of size 𝑚×𝑛, sampling frequency 𝑓_𝑠,

  duration 𝑇 second

(2) Output: The speech part of the signal 𝑄 of new size 1×𝑞_size,

  duration 𝑡

(3) for 𝑖=1 to 𝑛 do

         for 𝑗=1 to 𝑚 do

(4) 𝑦_𝑖=(𝑆_𝑖+𝑆_𝑗+1)/2

(5) end for

(6) end for

(7) initialize 𝐽=2

(8) 𝑆_smooth=median[𝑦(𝑖−(𝐽/2)),𝑦(𝑖−(𝐽/2)+1),…

…,𝑦(𝑖+(𝐽/2)−1)]

(9) 𝑧=𝑆_smooth/max(𝑆_smooth)

(10) Initialize 𝑇ℎ=0.04

(11) for 𝑖=1 to 𝑛 do

if |𝑧|>𝑇ℎ then 𝑞=𝑧

if |𝑧|≤𝑇ℎ then 𝑞=0

end if

end for

(12) Initialize 𝑄=0

(13) for 𝑖=1 to 𝑛 do

if 𝑞_𝑖≠0 then

𝑄=𝑞

else 𝑄=𝑄

end if

end for