在诊断一个有慢性咳嗽的病人时,他的咳嗽强度和频率评估能提供很有价值的信息.因此提高咳嗽识别率,对疾病的诊断有着重要意义.从语音识别中被广泛应用的Mel倒谱参数出发,寻找咳嗽和语音在Mel倒谱参数中的区别.基于Mel倒谱参数的原理,将其计算过程中的Mel刻度滤波器对数能量的极值数分布情况提取出来作为咳嗽的识别特征.在病房环境下对录音文件进行实验,得到的咳嗽识别率为90％以上,同时能够将语音等非咳嗽信号有效地剔除,实验结果显示90％以上的语音信号被排除.在录音设备及环境等各项参数不变的条件下,对不同病人样本,可使用同一阈值对咳嗽进行识别.该方法过程简单,数据计算量小,便于快速识别.%In the diagnosis of chronic cough patient, his cough intensity and frequency evaluation can provide valuable information. Therefore improving cough recognition rate has important significance for the diagnosis of disease. The article sets out from Mel-frequency cepstrum coefficient widely applied in the speech recognition, looking for the distinction of cough and voice in Mel-frequency cepstrum coefficient. Basing on the principle of Mel-frequency cepstrum coefficient, in the process of calculation the distribution of the extreme number of Mel scale filter logarithm energy are extracted as the recognition characteristics of cough signals. In the ward environment to the recorded files for experiment, the cough recognition rate is over 90% , at the same time voice and other non-cough signals are effectively removed, the experimental results show that more than 90% of the voice signals are ruled out. The recording equipment and environment and other parameters in the same conditions, for the different patient samples, the same threshold can be used to identify the cough. The method is simple, has small amount of data calculation, and facilitate quick identification.
展开▼
