A Study On The Characteristics Of Overnight Snoring Based On A Microphone Array-postural Model

Obstructive Sleep Apnea Hypopnea Syndrome is a systemic sleep apnea disease that is a major concern in society today.Different sleeping positions can change upper airway resistance,which affects the occurrence of sleep apnea and hypoventilation events.Currently,polysomnography is commonly used in clinical diagnosis of OSAHS.But this method is expensive and can cause discomfort to patients.Therefore,if we can obtain the patient's body position information through non-contact acoustic analysis method,it will be of great help to diagnose and study the pathology of OSAHS.First,we use the four-channel microphone array to collect all-night sleep sound signals,and pre-process the collected data and mark the sound segments.Then,we use the least squares method to extract the cross-correlation function of the array containing spatial information as features,and perform unsupervised clustering on the features by K-means.The class representative curve analysis of the array cross-correlation function got from the clustering result can obtain various body position information and the occurrence time of the body position,and the overnight body position changing curve can be obtained.The error rate from the actual infrared image is 8.8%.Finally,for the continuous and stable snoring signals under different body positions,we analyze the time-frequency spectrum diagram and the subjective perception of the human ear to learn that the time-domain waveforms,spectral energy distributions,and vocal tones of snore are different in different body positions.Besides,four major characteristic factors are extracted: peak frequency,center frequency,frequency center of gravity,and 800 Hz power ratio.It is found that the peak frequency and 800 Hz power ratio are greatly affected by the supine and non-supine positions,while the center frequency and the frequency center of gravity are basically not affected by any body position.It is concluded that the peak frequency and 800 Hz power ratio can be used as a good feature to identify the body position in subsequent studies.