| 摘要 |
A method and an apparatus for removing noise from an audio signal are provided to prevent deterioration of sound quality of an audio signal due to surrounding noise and change a channel gain only in an audio section to minimize sound distortion. An audio signal corresponding to a single frame is received and fast-Fourier-transformed(400,405). Channel energy is obtained from the Fourier-transformed signal for each frequency channel(410). A channel SNR(Signal-to-Noise Ratio) is calculated using the channel energy and noise energy(425), and an average of channel SNRs for respective frequency channels is calculated and determined as a spectrum distance(430). Long-term energy of a current frame is calculated using the channel energy and long-term energy of a previous frame(435), and a long-term SNR is calculated using the long-term energy and the noise energy(440). It is determined whether the frame corresponds to an audio section according to the spectrum distance(445). A channel gain of each frequency channel of the Fourier-transformed signal is calculated using the channel energy, the noise energy, and the channel SNR(455). When it is determined that the frame corresponds to the audio section, the channel gain is varied using the long-term SNR(465). The fast-Fourier-transformed signal is multiplied by the varied channel gain and inversely fast-Fourier-transformed(475). |
