Research On Recognition-based Detection Of Breathing Sounds During Overnight Sleep

Obstructive sleep apnea-hypopnea syndrome(OSAHS)is the most prevalent and most morbid disease in sleep respiratory disease.Common method for clinical diagnosis is to use a polysomnography(PSG)monitoring system,however,the PSG monitoring system is difficult to promote in homes and remote areas due to the complex and expensive equipment.The accurate diagnosis results of OSAHS require apnea events to be detected in a sleep record.Among the many simple diagnostic methods,screening based on acoustic signal analysis has become a research hotspot because of its simplicity and low cost.Therefore,the use of acoustic analysis methods to achieve snoring/breathing episodes(SBE)detection is very important to achieve OSAHS disease screening towards the family.This paper focuses on all-night sleep breathing sound signals,designs and implements an overnight SBE detection method based on pattern recognition.Firstly,the four existing SBE detection methods are analyzed,focused on the in-depth discussion and verification of sleep breathing sound detection methods based on artificial neural network(ANN).The results show that the detection performance of the method is significantly reduced in a low signal-to-noise ratio environment.In response to the above shortcomings,a dual ANN SBE detection method based on energy difference is proposed.This method divides the sleep breathing sound signal detection task into two based on the energy difference: detection of sleep breathing sound signals at low energy and snoring signal detection when energy is high.Finally,the results show that the accuracy of the overnight sleep sound detection method is 90.8%.On this basis,for the sleep sound signal collected in home environment,the signal-to-noise ratio is relatively low,and usually contains environmental noise and other human sounds.Further,the adaptive respiratory sound enhancement is realized by using a microphone array signal processing technology,and an artificial neural network based on energy difference is used to detect SBEs based on the enhanced SBEs,a method of all-night SBEs detection based on pattern recognition was finally constructed.The experimental results show that the accuracy rate of SBE detection by the above method is improved to 92.2%.The research results can effectively complete the task of detecting sleep and breathing all night,and have good medical value and social benefits.