SUMMARY

• The front-end converts speech data into observations vectors for statistical analysis by the speech recognizer.
  
  
• The Fourier transform, mean energy, digital filter banks, linear prediction, and the cepstrum are the standard digital signal processing (DSP) algorithms employed.
  
  
• Physiological knowledge of the human auditory and vocal articulatory systems is applied to better model speech and increase recognition performance.