欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
传统WDFT利用全通弯折函数AWF（AU—pass WarpingFunction）将单位圆上均匀分布的采样点，变换成非均匀分布的采样点。然而，这种计算方式会导致不同信号分量的延迟不同。本文基于匹配追踪的信号表示思想，提出了一种新的弯折傅里叶变换方法，实验结果表明，在原子个数大于信号长度情况下，该方法获得了较传统WDFT较好的弯折频谱表示性能。基于该方法表示的弯折离散傅里叶变换更适合于语音信号处理。
英文摘要：
Frequency axis warping to achieve nonuniform Fast Fourier Transform resolution uses a network of cascaded first order all-pass sections for frequency warping of the signal, followed by a standard FFT. However, different frequency component in the signal will have different signal delay and the all-pass filter is a IIR filter, so some high fre- quency component signal will be infinitude and will be truncated before DFT. Inspired by signal representation based on matching pursuit method, we propose the WDFT based on matching pursuit algorithm. The experimental results show a better warp spectrum representation when the number of atom is more than the length of signal. It is more suitable for perceptual speech processing than ordinary WDFT.
同期刊论文项目
同项目期刊论文
Multirate and Fast Parallel Implementation of 2-D DST Kernel-Based Real-Valued Discrete Gabor Transf
Blind Multiple Access Interference Suppression Algorithm Based on Relaxed Subgradient Projection for
Two-dimensional canonical correlation analysis and its application in small sample size face recogni
期刊信息
位置：