基于改进反向Mel频率倒谱系数的咳嗽干湿性自动分类
发布时间：2018-07-21 16:14
【摘要】：咳嗽的自动分类在临床上具有重要的辅助诊断作用。传统的Mel频率倒谱系数(MFCC)采用Mel均匀滤波器组,高频段的滤波器分布较稀疏,未能最大程度反映两类咳嗽的特征差别。针对这个问题,本文在分析干性咳嗽和湿性咳嗽频谱能量分布特点的基础上,提出了一种改进的反向MFCC提取方法,采用反向Mel刻度上的均匀滤波器组,并放置在两类咳嗽都具有高频谱能量的频段,使得特征提取集中在两类咳嗽特征信息丰富且差别显著的频段进行。基于隐马尔可夫模型的咳嗽干湿性自动分类实验结果表明,该方法获得了优于传统MFCC的分类性能,总体分类准确率从89.76%提高到了93.66%。
[Abstract]:Automatic classification of cough plays an important role in clinical diagnosis. The traditional Mel frequency cepstrum coefficient (MFCC) uses Mel uniform filter banks, and the filter distribution in high frequency band is sparse, which can not reflect the characteristic difference of two kinds of cough to the greatest extent. In order to solve this problem, based on the analysis of spectrum energy distribution characteristics of dry cough and wet cough, an improved reverse MFCC extraction method is proposed, which uses a uniform filter bank based on reverse Mel scale. And placed in the two kinds of cough have high frequency spectrum energy, so the feature extraction is concentrated in the two kinds of cough feature information rich and significant differences in frequency bands. The experimental results of automatic classification of cough dryness and wetness based on hidden Markov model show that this method has better classification performance than traditional MFCC, and the overall classification accuracy is improved from 89.76% to 93.66%.
【作者单位】： 电子科技大学中山学院机电工程学院;华南理工大学自动化科学与工程学院;广州医学院第一附属医院;
【基金】：中央高校基本科研专项基金项目资助(2012ZZ0106) 中山市科技计划项目资助(2014A2FC383)
【分类号】：R56;TP391.7
，
本文编号：2136075
[Abstract]:Automatic classification of cough plays an important role in clinical diagnosis. The traditional Mel frequency cepstrum coefficient (MFCC) uses Mel uniform filter banks, and the filter distribution in high frequency band is sparse, which can not reflect the characteristic difference of two kinds of cough to the greatest extent. In order to solve this problem, based on the analysis of spectrum energy distribution characteristics of dry cough and wet cough, an improved reverse MFCC extraction method is proposed, which uses a uniform filter bank based on reverse Mel scale. And placed in the two kinds of cough have high frequency spectrum energy, so the feature extraction is concentrated in the two kinds of cough feature information rich and significant differences in frequency bands. The experimental results of automatic classification of cough dryness and wetness based on hidden Markov model show that this method has better classification performance than traditional MFCC, and the overall classification accuracy is improved from 89.76% to 93.66%.
【作者单位】： 电子科技大学中山学院机电工程学院;华南理工大学自动化科学与工程学院;广州医学院第一附属医院;
【基金】：中央高校基本科研专项基金项目资助(2012ZZ0106) 中山市科技计划项目资助(2014A2FC383)
【分类号】：R56;TP391.7
，
本文编号：2136075
本文链接：https://www.wllwen.com/yixuelunwen/huxijib/2136075.html
最近更新
教材专著
