论文部分内容阅读
咳嗽是众多呼吸道疾病中常见的重要病症之一,其强度和发生频率提供了极其重要的临床信息。为利用这些信息,必须把咳嗽音从其他声音例如语音、清喉音、清鼻音等中分辨出来。我们提出一种基于经验模态分解(Empirical Mode Decomposition,EMD)分析的咳嗽音检测方法。该方法通过应用EMD的自适应滤波器组特性,提取信号的频域能量分布以统计分析咳嗽音及语音等特征,进而找到优化特征提取的方法,并利用隐马尔可夫模型(Hidden Markov model,HMM)进行咳嗽音的检测。临床数据的实验表明,该优化方法能有效提高咳嗽音检测的正确率。
Cough is one of the most common diseases in many respiratory diseases, and its intensity and frequency of occurrence provide extremely important clinical information. In order to utilize this information, it is necessary to distinguish the cough sound from other sounds such as voice, clear throat sounds, clear nasal sounds, and the like. We propose a cough tone detection method based on Empirical Mode Decomposition (EMD) analysis. This method uses the adaptive filter bank of EMD to extract the energy distribution of the signal in the frequency domain to statistically analyze the characteristics of cough sounds and speech, and then find out the method to optimize the feature extraction. Using Hidden Markov model (Hidden Markov model) HMM) for the detection of cough sounds. The experiment of clinical data shows that this optimization method can effectively improve the correctness of cough tone detection.