摘要
对具有不同类型时频结构的复杂信号,为了最优化非线性信号逼近,可根据信号自适应地选择基。分析了通过极小化凹花费函数,从基字典中挑选“最佳”基的原理,采用快速最佳基的树搜索算法,在小波包基和局部余弦基这一类基中寻找被处理信号的最优基,实现了含噪语音录音信号的去噪。结果表明,最优经验局部余弦基对此类复杂信号的去噪效果远比固定小波基阈值去噪效果好。
For complex signal with different time-frequency frame, the best basis adaptive signal can be selected for optimization of non-linear approximation. This paper analyzes the theory of ＂best＂ basis selection from a library of bases by minimum estimation of risk. The fast best-basis algorithm is used to pick out the ＂best＂ basis from all wavelet packet basis and local cosine basis, to actualize denoising of speech recording signal obtained by adding a Gaussian white noise. The simulation shows that the estimated signal recovered from the local cosine coefficients above the threshold in the best basis is far better than from appointed wavelet basis.
出处
《西安科技大学学报》
CAS
北大核心
2007年第2期296-299,共4页
Journal of Xi’an University of Science and Technology
基金
陕西省自然科学基金资助项目(2004F30)
关键词
非线性逼近
自适应基
阈值去噪
局部余弦基
non linear approximation
adaptive base
threshold denoising
best local cosine basis
