Breath Sounds Feature Extraction And Classification And Identification Method

Respiratory sounds is a kind of acoustic signal in the ventilation process of human respiratory system with external environment, and it contains much information of pathology and physiology. However, because of its randomicity, variations in the details of sound capture and analysis techniques between researchers, it makes respiratory sounds study results greatly different and comparison of results from the different centers difficult. With the development of computer and signal processing technique, respiratory sounds recognition has become a new focus and it is worth paying more and more attention. Feature extraction is the core of pattern recognition. The extracting process is to make effective features appear in the transform domain by transform method and throw off the useless information and look for the most efficient signal feature to form pattern feature vectors for classification.For a steady respiratory sounds analysis system, its transfer function remains and the corresponding amplitude-frequency and phase-frequency characteristics are fixed. So the sampled data can reflect the physical characteristic of signal to some degree.It is found that the waveforms in time domain and in transform domain of different respiratory sound signals are different. The difference can be described by envelope which can show the characteristics of signals. And it needs to be combined with filters with conventional methods (Hilbert transform, for example). So a novel method of respiratory sounds envelope extraction based on complex wavelet transform is put forward in this paper.After preprocessing of collected respiratory sounds data firstly, four kinds of typical respiratory sound signals (normal, tracheitis, pneumonia and asthma) are sampled from various subjects. By studying the time-frequency distribution characteristic of respiratory sounds, we select Morlet wavelet and proper scales and do continuous wavelet transform (CWT) to obtain envelope. A set of statistical features and energy of envelope are extracted to form the input feature vectors for artificial neural network (ANN). And the recognition rate reaches 91.7%.Respiratory sounds are composed of multi-frequency components and take on obvious periodic wave. Normal and abnormal respiratory sounds are not the same in...