According to one embodiment, various feature parameters are calculated for distinguishing between a speech and music and between music and background sound for an input audio signal. With the feature parameters, score determination is made as to whether the input audio signal is close to a speech signal or a music signal. If the input audio signal is determined to be close to music, the preceding score determination result is corrected considering the influence of background sound. Based on the corrected score value, a sound quality correction process for a speech or music is applied to the input audio signal.