• speech segment is weighted by Hamming window
  
  
• fast Fourier transform of windowed speech segment is computed to transform it into frequency domain
  
  
• real and imaginary parts of short-term speech spectrum are squared and added to get the short-term power spectrum