一种新呼吸音信号特征提取方法与应用

目的:通过研究呼吸音信号与呼吸系统疾病内在联系，对呼吸音特征进行提取与分类，为研制便携家用型呼吸音远程监测移动装置进行技术准备。 方法:对采集到的呼吸音进行预处理分析，从处理后的呼吸音原始数据中提取短时能量与短时过零率的特征值。 结果:通过不同时间段呼吸音信号能量的高低来显示特征差异，并且呼吸音的高低频异常信号对特征提取方法影响较小。 结论:本文方法可简单有效地提取特征值，不仅简化了特征识别数据处理过程，而且提取到的特征参数满足了差异性、统一性及相关性等基本特征。为构建神经网络的输入提供了理论依据与实际数据支撑。

A new feature extraction method of respiration signal and its application

Objective To extract and classify the respiratory sound characteristics by studying the internal relationship between respiratory sound signals and respiratory diseases and provide technical preparations for developing a portable household breathing sound remote monitoring mobile device. Methods After the pretreatment analysis of the collected respiratory sounds, we extracted the characteristic values of short-time energy and short-time zero crossing rate from the processed respiratory sound original data. Results The characteristic differences were displayed by the energy changes of different time quantum respiratory signals, and the results indicated that the high-and low-frequency abnormal signals of respiratory sounds have trivial influence on the feature extraction method. Conclusion The proposed method can be used to extract characteristic values simply and effectively, not only simplifying the characteristics recognition and data processing, but also obtaining the extracted feature parameters which meet the basic characteristics, including otherness, uniformity and correlation. The study provides a theoretical basis and actual data support for establishing the input of neural network.