摘要 |
PURPOSE: A system and a method for improving voice are provided to extract a voice signal from which noise is removed, from an input frame by using only time area, thereby realizing simple constitution. CONSTITUTION: A parameter estimating part(201) estimates predetermined parameters required for filtering, from an input frame in real time. A sample detector(211) detects a sample related to the input frame. A gain matrix detector(212) detects a gain matrix of a filter(210) to a voice signal included in the input frame by using the estimated parameters. A first estimation detector(213) detects an estimation of a state variable of the voice signal by using the detected sample and gain matrix. A second estimation detector(214) detects an estimation of the voice signal from which a noise signal is removed, from the estimation detected in the first estimation detector.
|