摘要 |
PROBLEM TO BE SOLVED: To provide a small-sized and inexpensive voice extraction controller capable of proper voice extraction by means of common components without having two voice extraction constitutions when accompaniment of plural voice signals is identical and not, in a device of extracting voice signals that are a model for voice training. SOLUTION: Signals output from a gain control means 23 capable of changing the polarity and amplitude of voice input signals and the other voice input signals are added, and model singing suitable to the voice input signals is extracted by controlling gains according to the results of which the outputs are bandwidth-restricted and the results of which the amplitude and phase of each of the results of which the other voice input signal is bandwidth- restricted is measured. |