To provide a device of less quality deterioration by correcting a parameter value from values of a formant frequency, band width, amplitude and a basic frequency, etc., when the processing of voice quality conversion, etc., is performed by using formant analytic synthesis.
A voice signal inputted for converting voice quality is converted to an electric signal by a voice input part 1, and the voice signal making a low band signal a main among the signals is stored in a signal processing buffer 5 to be processed in a frame unit. In the processing, a pitch extraction part 7 calculates a pitch frequency at every frame, and a power calculation part 6 calculates power of respective frames, and a formant analytic part 8 analyzes the characteristic amounts of the formant frequency and the band width, etc., at every frame. A V/UV decision part 12 decides V(voice)/UV(voiceless)/S (soundless) by using the obtained power and pitch parameters. A parameter correction part 14 corrects the parameter when the probability that the extracted parameter causes abnormal amplitude exists.
MATSUI KENJI
Next Patent: COMMUNICATION TERMINAL