摘要 |
PROBLEM TO BE SOLVED: To make it possible to easily stress the Formant of speech spectra faithfully with original speeches by reproducing the converted speeches weighted in a frequency region. SOLUTION: The speech intermediate output signal yi (i=1 to n) obtained by a band-pass filter group 5 [Fi (i=1 to n)] with respect to an input speech signal (x) is transmitted to an EIH spectrum analyzer 3, a reference spectrum analyzer 6 and a modulator 7. The EIH spectrum analyzer 3 outputs EIHi and the reference spectrum analyzer 6 outputs a logarithmic value SOi of the short time mean intensity of the speech intermediate output signal yi. The EIHi and SOi are sent to a weight calculation circuit 7-1 in the modulator 7, from which a weight coefficientγi is outputted. The weight coefficientγi is sent to an amplitude modulator 7-2 where the modulation to weight the speech intermediate output signal yi is executed to obtain the weighted modulation output signal wi. The modulation output signal wi is sent to an adder 4 and is subjected to addition, by which a stress output signal (z) is obtained.
|