| 摘要 |
The method involves sampling a speech signal and then segmenting the signal into sub-bands. The resultant signal in each sub-band is transposed into a baseband signal (2). A calculation is then performed to determine the square of the auto-correlation modulus (p(M)) of the transposed signal in each sub-band. The fundamental frequency value of the speech signal is estimated by selection from amongst those auto-correlation signals which have the maximum amplitude. Selection is then performed (4) in each maximum auto-correlation sub-band which corresponds to a value at the chosen fundamental frequency and at multiples and sub-multiples of this frequency. These signals are then corrected prior to transmission.
|
