摘要
基于db小波包变换,采用频率分级阈值方法对三维荧光光谱数据进行了压缩。建立了数据的小波包分解树,根据对数能量熵最小原则确定最优树,通过频率分级阈值方法对最优树中的小波包系数进行压缩,并且用实验获取的数据加以验证。实验结果表明,和小波变换相比小波包变换能够更有效地保留数据的细节信息。通过和其他阈值法比较可知,频率分级阈值法具有更好的压缩率和数据恢复能力,其压缩分数达到90%,恢复分数大于98%,谱线相对误差小于1%。平行因子分析重构光谱与原始光谱的结果表明,小波包压缩能有效保留有用信息。
Excitation-emission- matrix （EEMs） data have been compressed with a frequency grade threshold based on a Daubechies （db） wavelet packet transform. First, the tree of the wavelet packet was formulated; sec- ondly, the best-tree was determined by the principle of minima; finally, the frequency grade threshold was used to compress the wavelet packet coefficient. Experimental data were used to test the compression ability of this method. The compression results indicated that wavelet packet compression could retain more information than wavelet com- pression. The frequency grade threshold method based on wavelet packets was adopted to compress EEMs. The fre- quency grade threshold method afforded higher reconstruction percentages and compression percentages than other threshold compression methods： the maximum reconstruction percentage reached 90% , and the compression per- centage was larger than 98% , with errors between spectra of lower than 1%. With the use of parallel factor analysis （ PARAFAC ） of the original EEMs data and the compressed data, it was possible to obtain the characteristics of the excitation and emission spectra of each component. The results of PARAFAC show that wavelet packet transform can compress and recover the information in EEMs data effectively.
出处
《北京化工大学学报（自然科学版）》
CAS
CSCD
北大核心
2013年第6期100-105,共6页
Journal of Beijing University of Chemical Technology(Natural Science Edition)
基金
国家自然科学基金(50975019)
关键词
数据压缩
三维荧光光谱
小波包变换
平行因子法
compression
fluorescence analysis
wavelet packet
parallel factor analysis
