Research On The Characteristics Of Snoring In Supine Position Based On Microphone Array

Obstructive Sleep Apnea Hypopnea Syndrome(OSAHS)is a common sleep breathing disorder,and loud snoring is the most significant symptom in patients.Snoring can induce a series of cardiovascular diseases,and can affect the sleep of the sleeper in the same room,easily leading to interpersonal conflicts.Current medical diagnoses of OSAHS patients often use Polysomnography(PSG),which is expensive and can cause "invasive" discomfort.Some studies have shown that snoring signal contains related pathological information such as upper airway tissue vibration,and in recent years,non-contact snoring diagnostic method based on acoustic monitoring has become one of the research hotspots.Because changes in sleep position can affect the patient's upper airway resistance and thus change the characteristics of snoring,it is of great significance for non-contact pathology diagnosis in OSAHS patients to analyze snoring characteristics in combination with different positional states.Considering that patients' snoring onset rate is the highest in supine condition,this thesis,based on the National Natural Science Foundation of China(61271410)general project,focused on the characteristics of snoring in supine position based on microphone array.Firstly,the least square cross-correlation function was proposed as the spatial characteristics of snoring,and the unsupervised clustering method of K-means was used to distinguish patients' supine and non-supine positions.The accuracy rate reached 90%,and the spatial characteristics of snoring under the two positions were analyzed.Secondly,in view of the trachea in the process of inspiratory and expiratory change,the pathological information carried by different phenomena,study a snoring classification method based on acoustic features fusion.Sleep snoring extracted waveform under the supine position,frequency and energy,formant acoustic features,to incorporate all characteristics through Relief algorithm selection,Twin support vector machine(TWSVM)was used to classify snoring in supine position into inhaling snoring and exhaling snoring,and the fusion feature classification accuracy reached 92.4%,which was much better than the single feature classification ability.Finally,based on the sleep snore data of the ward all night,the experimental verification was carried out to realize the discrimination of sleep position and the classification of snoring under supine position.At the same time,the characteristics of snoring data of supine position all night were studied,and it was found that there were significant changes in the frequency characteristics of snoring under supine position at different sleep stages.The work in this thesis lays a foundation for the further study of the postural-snore characteristic model,which has good medical value and social benefits.