Conclusion:
  • Advantages :
    • Easier to code.
    keep the pitch of the speaker fairly constant.
  • Disadvantages:
    • Difficult to analyse.
    Not very robost.
    Not very useful for real time applications.


    Observations:

     
    Total Number of Samples
    Original signal
    No Overlap
    33% Overlap
    50% Overlap
    66% Overlap
    59802
    76926
    83407
    90884
    106817
    Reason :
  • As we decrease the overlap, we find that the number of windows increases. As we find the optimum shift for each window, the total optimum shift for the entire speech signal, increases.
  • The number of samples to be considered to determine the optimum shift for each window also, varies from window to window.
  • The discontinuties in the warped speech, is due to the mismatch.