Robustness of signal correlation
Wee-Soon Ching, Peng-Seng Toh · 2002
The robustness of signal correlation is of paramount importance if signal correlation is to be used effectively in the real world. The paper examines the robustness of speech correlation based on the average correlation between adjacent speech frames. The effects of the speech frame size, the amount of amplitude distortion and the amount of gaussian noise on the correlation property are studied. Simulation results show that the optimum speech analysis frame size is in the region of 15 to 20 msec. The speech correlation property is found to be robust even when the amplitude of the speech signal is infinitely clipped. It can also withstand fairly high gaussian noise and has numerous applications in the field of speech coding, speech enhancement and speech analysis.>