利用腰带式多参数生理信号监测系统进行咳嗽检测与辨识

Detection and Identification of Cough by Belt-style Multi-parameter Physiological Signals Monitoring System

目的:以腰带式多参数生理信号监测系统为实验平台,提出一种通过实时记录咳嗽引起的腹部加速度信号、振动信号和咳嗽声音来监测咳嗽的新方法,并针对后期数据的回放处理设计一种能够自动识别咳嗽事件的算法。方法:将采集到的数据用MATLAB在计算机上进行分析:根据加速度信号的幅度和斜率,可以排除说话的干扰;再利用呼吸波信号的幅度,筛除掉清喉事件;最后由声音包络的变化识别出咳嗽事件。结果:该方法能够有效区分咳嗽与其他干扰(如清喉、说话和体动):对同步记录的音频文件进行人工计数,共有523次咳嗽事件,算法自动识别出471次,咳嗽事件识别的敏感度为90.1%;共有1 452次非咳嗽事件,算法正确识别出1 438次,特异性达到99%。结论:利用腰带式多参数生理信号监测系统进行咳嗽检测与辨识的方法是有效可行的,可以推广应用。

Objective To use belt-style multi-parameter physiological signals monitoring system as the experiment platform to put forward a new method for monitoring cough by real-time recording of acceleration signal, vibration signal and sound of cough, and to design an automatic cough event recognition algorithm for playback of late data. Methods The acquired data were analyzed by MATLAB. The amplitude and slope of acceleration signal were used to eliminate the interference from uttering. The amplitude of respiratory wave signal was used to screen the influence of clearing throat event. Cough event was recognized with the change of sound envelope. Results The method could distinguish between cough and other interferences such as clearing throat, uttering and body movement. Artificial counting of simultaneously recorded audio files was performed. There were 471 times of cough events recognized by the algorithm from 523 ones, with a sensitivity of 90.1%; there were 1 438 times of non-cough events recognized by the algorithm from 1 452 ones, with a sensitivity of 99%. Conclusion Belt-style multi-parameter physiological signals monitoring system can be used for the detection and recognition of cough, and thus is worth popularizing clinically.