Keywords:
Equalization filter; Speech signal; Speech noise reduction; Frequency band overlap
均衡滤波器;语音信号;语音降噪;频带重叠
Abstract:
Objective: To explore the overlap between adjacent frequency bands in equalization filters, and an audio
clipping and splicing algorithm based on Fourier transform is proposed. Methods: Aiming at the speech collected in the
real scene, noise reduction is carried out on the graph filter based on Fourier transform. The audio processing effects
before and after using audio clipping and splicing algorithm are compared, and the speech waveform diagram and
spectrogram are drawn. Results: In the filter with audio clipping and splicing algorithm, the problem of overlapping
between adjacent frequency bands can be effectively solved and the pulse disturbance can be avoided. Conclusions: The
audio clipping and splicing algorithm based on Fourier transform can effectively solve the overlapping problem between
adjacent frequency bands of equalization filter, and can be used in many audio processing scenarios such as speech noise
reduction and speech enhancement.
目的:探究均衡滤波器中相邻频带间的重叠问题,并提出了一种基于傅里叶变换的音频剪切和拼接算法。方法:针对真实场景下采集的语音,在基于傅里叶变换的均衡滤波器上进行降噪处理,分别对比了采用音频剪切和拼接算法前后的音频处理效果,绘制了语音波形图和语谱图。结果:在采用了音频剪切和拼接算法的滤波器上,能有效解决相邻频带间的重叠问题,避免了脉冲扰动。结论:基于傅里叶变换的音频剪切和拼接算法能有效解决均衡滤波器相邻频带间的重叠问题,可用于语音降噪、语音增强等多个音频处理场景。