CQT and STFT deep speech spectrum feature based snore classification method and system

Abstract

The invention relates to medical instruments, in particular to snore classification, and aims to realize automatic snore recognition. According to the technical scheme, a CQT (constant Q transformation) and STFT (short-time Fourier transformation) deep speech spectrum feature based snore classification method is provided; CQT and STFT are carried out on snore audio signals, a speech spectrogram generated after transformation serves as an input of a pre-trained deep convolutional neural network, an output of the pre-trained deep convolutional neural network is extracted to serve as a feature vector, a classification model is trained by using an SVM (support vector machine), and finally automatic snore recognition is realized by using the trained classification model. The snore classification method is mainly applied to design and manufacturing occasions of snore classification medical instruments.

Description

technical field [0001] The present invention relates to medical equipment and snoring sound classification, in particular, to a snoring sound classification method and system based on CQT and STFT depth spectrum features. Background technique [0002] Snoring is a sign of weakened breathing during sleep. 20% of people snore, and 15% of snorers suffer from Obstructive Sleep Apnea (OSA) syndrome. The mortality rate is as high as 40%. There are about 3,750 in our country. The health of thousands of people is threatened by the disease. OSA is a common sleep disorder characterized by repeated apnea and snoring (ie, "snoring") during sleep, clinically referred to as "snoring". The pathogenesis of obstructive sleep apnea, which has been investigated for 25 years, is essentially due to the narrowing of the upper airway or the relaxation and collapse of soft tissues during sleep, resulting in partial or complete obstruction of the upper airway; resulting in low levels of sleep durin...

AI Technical Summary

Problems solved by technology

First, it requires specialists for safe administration and monitoring of sedation; second, it is time-consuming, typically taking 20 minutes for a single examination; moreover, it cannot be performed during the patient's natural sleep, and the invasive endoscopy can make it difficult for patients to fall asleep or wake up from sleep

Time-consuming and laborious, and the current research on snoring blockage and vibration location based on snoring signals is still in its infancy, and the accuracy needs to be improved

Images