| 摘要 |
<P>PROBLEM TO BE SOLVED: To provide a device and method for extraction of a vocal tract cross section function, capable of excellently controlling voice quality even for a voice signal of a high sampling frequency. <P>SOLUTION: The extraction device of the vocal tract cross section function extracts the vocal tract cross section function from a digitalized voice signal. The device includes: a frame segmenting unit 10 for segmenting the digitalized voice signal into frame units; a band divider 20 for dividing each voice signal into a low frequency component and a high frequency component; a flattening processing unit 30 for flattening a frequency characteristic of each component; an autocorrelation function conversion unit 40 for converting each component into an autocorrelation function; an LSP (Line Spectrum Pair) parameter extraction unit 50 for extracting the LSP parameter from the autocorrelation function of each component; an LSP parameter connection 60 for connecting the LSP parameter of each component; a PARCOR (Partial Autocorrelation Coefficients)coefficient conversion unit 70 for converting the connected LSP parameter into a PARCOR coefficient; and vocal tract cross section function conversion unit 80 for converting the PARCOR coefficient into the vocal cross section function. <P>COPYRIGHT: (C)2011,JPO&INPIT |
