摘要
我国电影已经有一个多世纪的历史了,有些老电影音频的感光膜在岁月的腐蚀和人为的损伤下,会出现霉斑、划痕等痕迹,在还音过程中会产生背景噪音,从而影响听感。对此,提出一种基于SE-GAN的音频修复算法,能有效地解决老电影的音频质量问题。该算法通过引入矢量量化层来提高信号发生器的性能,引入频域损耗,使得去噪后的语音信号具有更接近于纯净音频的频率特性。试验证明,该算法能有效地改善老电影的去噪性能,并使其各项客观评价指标得到改善。
Chinese movies have a history of over a century.Some old movie audio films have sensitive films that,under the corrosion of time and human damage,may show signs such as mold and scratches.During the sound reproduction process,background noise is produced,which affects the listening experience.Based on this,this article proposes an audio restoration algorithm based on SE-GAN,which can effectively solve the sound quality problem of old movies.This algorithm improves the performance of the signal generator by introducing a vector quantization layer,and introduces frequency domain loss to make the denoised speech signal have frequency characteristics closer to pure audio.Experiments have shown that this algorithm can effectively improve the denoising performance of old movies and improve their various objective evaluation indicators.
作者
丁雄
DING Xiong(Aiyin Software Technology Co.,Ltd.,Beijing 100012,China)
出处
《电声技术》
2024年第1期29-31,共3页
Audio Engineering
关键词
老电影
音频修复
音频降噪
old movies
audio repair
audio noise reduction
