摘要 |
PROBLEM TO BE SOLVED: To provide a recorded voice clarifying device capable of improving clarity stepwise for a recorded voice which is recorded via a partition.SOLUTION: Object voice analyzing means 10 performs frequency analysis of an object voice signal and calculates an object voice intensity spectrum S(f, τ) defined as a variable f and a frame τ based on the frequency. Reference voice analyzing means 20 performs frequency analysis of a reference voice signal and calculates a reference voice average value spectrum Hav(f) which is a time direction average. Noise component spectrum preparing means 30 prepares a noise component spectrum N(f) using Sav(f). Modulation component spectrum preparing means 40 prepares a modulation component spectrum G(f) by dividing Hav(f) with a value subtracting N(f) from Sav(f), and voice signal correction means 50 subtracts N(f) by a ratio α for S(f, τ), and a correction voice signal is obtained by dividing G(f) by a ratio &bgr;. |