Snore detection method and device based on time frequency similarity

Abstract

The invention discloses a snore detection method and device based on time frequency similarity. The method includes the following steps that collected snore audio signals are acquired, and after the snore audio signals are subjected to noise estimation, voiced fragment extraction is carried out through the end point detection technology; the Euclidean distance of time domain energy or frequency domain energy of a current voiced fragment and a former voiced fragment is calculated, the Euclidean distance is compared with a threshold value, and the snore state is judged according to the comparison result. Average frequency spectrum energy and time domain energy of the voiced fragments serve as characteristics, the time frequency similarity of two adjacent voiced fragments is used for snore detection, the operation cost is reduced, and the requirements of hardware implementation and real-time detection are met. The short-time change of snores is neglected, the snores are detected starting from the periodicity of the snores within a long time, snore data does not need to be well labeled, and the method and device can adapt to most snore conditions.

Description

technical field [0001] The invention belongs to the technical field of sound signal detection, and in particular relates to a snore detection method and device based on time-frequency similarity. Background technique [0002] The incidence rate of obstructive apnea syndrome (OSAHS) is about 3%-4%, which is called "snoring" clinically. OSAHS seriously affects the quality of life of patients, and can easily lead to impaired learning and memory, low work efficiency, frequent traffic accidents, and disorders of the endocrine and endothelial systems. Therefore, the detection of the snoring signal is very important for the tracking of the sleep process and the judgment of the sleep quality. [0003] However, the existing snoring signal detection generally utilizes a polysomnography monitoring system, which is complex and expensive, and "invasive" monitoring can easily cause physical discomfort. Among the many simple diagnostic methods for OSAHS, there are many related studies us...

AI Technical Summary

Problems solved by technology

In addition, the existing snoring detection technology has the following disadvantages when processing recording data in real scenes: 1. It depends on the validity of features and the reliability of classification models, and the extraction process of common features in the prior art is very complicated. The computational cost is high, and it is difficult to meet the needs of hardware implementation and real-time detection; the reliability of the classification model depends on a large amount of snoring data with well-labeled information, which is difficult to obtain in real scenes; 2. The method of extracting features frame by frame Ignoring the periodicity of snoring in a longer time span, the effectiveness of the method is difficult to generalize to most snoring

Images