摘要
在需要对信号进行再量化的场合,可以通过加入dither来避免小信号再量化所产生的谐波失真,但同时会使噪声功率增加。这种情况下,可以利用人耳的心理声学特性,通过噪声整形来降低噪声的可闻性,提高实际的信噪比,改善音质。本文提出了两种新的设计最优噪声整形滤波器的方法-遗传算法和非线性优化算法,并分别实现了原采样率下和过采样率下基于心理声学模型的最优噪声整形滤波的设计。结果证明,该方法灵活方便、实现效果良好。
In the occasion that signal needed to be re-quantified, adding proper dither can avoid the harmonic distortion generated in small signal＇s re-quantization, but it will make noise power increase at the same time. In this case, we can take advantage of human ear＇s psychoacoustic characteristics, apply noise shaping technology to reduce the noise＇s audibility, then improve actual signal to noise ratio （SNR） and improve sound quality. This paper presents two new methods to design optimal noise shaping filters-genetic algorithm and nonlinear optimization algorithm, and realized the design of optimal noise shaping filters based on psychoacoustic model respectively under original sampling rate and over-sampling rate. It proves that these methods are flexible and convenient, can achieve good results.
出处
《电子设计工程》
2013年第4期75-78,共4页
Electronic Design Engineering
基金
陕西省教育厅专项科研计划项目(11JK0548)
