Snoring Sounds Research Articles

Ensemble Heterogeneous Feature Selection (EHFS) and Heterogeneous Ensemble Classifier with VOTE (HECV) for Automatic Detection of Snore Sounds

Obstructive Sleep Apnea (OSA) is generally considered as a sleep co-related breathing difficulty with some important well known disabling indication. This research work tends to differentiate the condition of OSA victims. The significant work is to propose an innovative Heterogeneous Ensemble Classifier with Velum Oropharyngeal Tongue Epiglottis (HECV) method with a high level of effectiveness. In the formulated study, data recording, features extraction, multifeature selection, classification and performance evaluation are the five stages of processing. At the initial stage, the noises represented in the audio signals which were eliminated through Adaptive Fuzzy Median Filter (AFMF) algorithm. After thatCrest Factor, original Frequency, Spectral Frequency Features, Subband Energy Ratio, Mel-Scale Frequency Cepstral Coefficients (MFCC), Empirical Mode Decomposition (EMD) Features, and Wavelet Energy Features are collected from the noise suppressed audio signals and inputs are fed into Ensemble Heterogeneous Feature Selection (EHFS) technique. EHFS algorithm fuses the outputs of filter and wrapper oriented feature selection methods. These identified features are classified using HECV technique which renders a good classification outputs by validation without the regard of subjects. The outputs show that the formulated HECV approach gives better performance in snore detection when co-related with other classifiers.

Smart Phone based Snoring Sound analysis to Identify Upper Airway Obstructions.

Obstructive sleep apnea (OSA) is characterized by upper airway obstructions known as apnea/hypopnea events. Narrowing of the upper airway during or near the vicinity of apnea/hypopnea causes the spectrum of the snores to shift to higher frequencies. Using an instrumentation quality wideband (WB) microphone (4Hz-100kHz), we previously demonstrated that potentially diagnostically useful frequency shifts could be detected even in regions beyond the human hearing range. WB-microphone based systems are expensive and not available for home use or population screening application. In this paper we explore the feasibility of using smart phones to analyze snoring sounds in the 20Hz-22kHz band to identify events of upper airway obstructions. Modern smart phones have internal microphones with bandwidths up to 22kHz, above the nominal human hearing range, and provide a good platform for sound acquisition and processing. For the work of this paper we used a Samsung Galaxy S3 phone and recorded overnight respiratory sound data from 8 patients undergoing routine Polysomnography (PSG) study in a hospital. Our target was to develop models to classify each standard 30 second epoch of data as non-apnea or apnea. Using 700 epochs we developed logistic regression models with the input as snoring sound features and the outputs as the diagnostic classification of each event (apnea/non-apnea). Models developed within a 20Hz-15kHz band had accuracies of 89-93%, sensitivities 70-78% and kappa index ranging 0.75-0.83 on validation data set. When the same models were developed on the 20Hz-22kHz frequency band the improved performance shows accuracies 94- 97%, sensitivities 93-100%, and kappa ranging 0.86-0.91. The study shows that smart phones based high frequency band (15-22kHz) of snoring sounds carry information about the upper airway obstructions. Our non-contact, smart phone based snoring sound technology has potential to identify upper airway obstructions.

Snoring - An Acoustic Definition.

Objective- The distinction of snoring and loud breathing is often subjective and lies in the ear of the beholder. The aim of this study is to identify and assess acoustic features with a high suitability to distinguish these two classes of sound, in order to facilitate an objective definition of snoring based on acoustic parameters. Methods- A corpus of snore and breath sounds from 23 subjects has been used that were classified by 25 human raters. Using the openSMILE feature extractor, 6 373 acoustic features have been evaluated for their selectivity comparing SVM classification, logistic regression, and the recall of each single feature. Results- Most selective single features were several statistical functionals of the first and second mel frequency spectrum-generated perceptual linear predictive (PLP) cepstral coefficient with an unweighted average recall (UAR) of up to 93.8%. The best performing feature sets were low level descriptors (LLDs), derivatives and statistical functionals based on fast Fourier transformation (FFT), with a UAR of 93.0%, and on the summed mel frequency spectrum-generated PLP cepstral coefficients, with a UAR of 92.2% using SVM classification. Compared to SVM classification, logistic regression did not show considerable differences in classification performance. Conclusion- It could be shown that snoring and loud breathing can be distinguished by robust acoustic features. The findings might serve as a guidance to find a consensus for an objective definition of snoring compared to loud breathing.

DETECTING NONLINEAR AND NONSTATIONARY PROPERTIES OF POST-APNEIC SNORING SOUNDS USING HILBERT–HUANG TRANSFORM

This study focuses on patients with severe obstructive sleep apnea syndrome (OSAS), and clarifies the existence of nonlinear and nonstationary properties in post-apneic snoring sounds. Many researchers have tried to discover intrinsic properties of the snoring sounds in OSAS patients for the past decades using linear frequency analysis, but no one has shown any evidence of the existence of nonlinearity and nonstationarity based on the quantitative evaluation of the post-apneic snoring sounds. In this study, Hilbert–Huang transform (HHT), which is designed for analyzing nonlinear and nonstationary signals, is adopted to generate a time-frequency map and the temporal variation in the spectral density is quantitatively evaluated using the averaged Kullback–Leibler divergence (AKL). As a result, for six OSAS patients, there is a tendency that most of AKL calculated from post-apneic snores are higher than those from non-apneic snores, which indicates that post-apneic snores are more nonstationary. In addition, we also evaluated the difference between the HHT time-frequency maps and spectrograms using short-time fourier transforms (STFT). Such analyses revealed that frequency fluctuations inherent to snoring can be adequately represented through HHT, but not with STFT. These nonlinear and nonstationary properties seem to be highly related to the physiological phenomena of OSAS, two of which are the explosive airflow after reopening of the closed airway and the collision vibration of the soft tissues.

Snore-GANs: Improving Automatic Snore Sound Classification With Synthesized Data.

One of the frontier issues that severely hamper the development of automatic snore sound classification (ASSC) associates to the lack of sufficient supervised training data. To cope with this problem, we propose a novel data augmentation approach based on semi-supervised conditional generative adversarial networks (scGANs), which aims to automatically learn a mapping strategy from a random noise space to original data distribution. The proposed approach has the capability of well synthesizing "realistic" high-dimensional data, while requiring no additional annotation process. To handle the mode collapse problem of GANs, we further introduce an ensemble strategy to enhance the diversity of the generated data. The systematic experiments conducted on a widely used Munich-Passau snore sound corpus demonstrate that the scGANs-based systems can remarkably outperform other classic data augmentation systems, and are also competitive to other recently reported systems for ASSC.

Effects of Increased Pharyngeal Tissue Mass Due to Fluid Accumulation in the Neck on the Acoustic Features of Snoring Sounds in Men.

Snoring sounds are generated by the vibration of pharyngeal tissue due to the upper airway narrowing. While recorded by a microphone placed over the neck, snoring can pass through the pharyngeal tissue surrounding the upper airway. Thus, changes in the pharyngeal tissue content may change the acoustic properties of the snoring sounds. Rostral fluid shift and the consequent increases in neck fluid volume (NFV) and neck circumference (NC) can increase pharyngeal tissue mass. Therefore, the goal of this study was to investigate the relationship between increases in pharyngeal tissue mass, as assessed by increased NFV and NC, and snoring sounds features. We obtained data from a previous study where 20 males who were not obese participated in a daytime polysomnography and their NC and NFV were measured before and after sleep. During sleep, snoring sounds were recorded with a microphone placed over the neck. Spectral centroid of the snoring sounds was estimated. Then, the first five snoring segments were selected from the first and last 30 minutes of stage N2 sleep. We found a significant decrease in the snoring spectral centroid from the beginning to end of sleep. We also found that spectral centroid from the end of sleep in frequency ranges below 200 Hz was inversely correlated with the increases in NFV and NC from before to after sleep. These results suggest that snoring spectral centroid can be used as a noninvasive and convenient method to assess variations in the pharyngeal tissue mass.

Identifying upper airway obstructions using wide band analysis of snore sounds

Identifying upper airway obstructions using wide band analysis of snore sounds

Characterisation of upper airway obstructions using wide-band snoring sounds

Background: Obstructive sleep apnea (OSA) is associated with partial (hypopnea) and full (apnea) upper airway obstructions. The standard method for identifying these obstructions and leading to OSA diagnosis is overnight polysomnography (PSG). One of the important outcomes of PSG is the apnea-hypopnea index (AHI), which represents the average number of obstructions per hour, computed over the total duration of sleep. AHI is a single number summary of the dynamics of the upper airway during sleep. PSG is complicated, requires more than 20 body-contact sensors and is not available for population screening. There is a need for simple and automated technologies that can characterise the upper airways in more detail. In this paper, we address this need by proposing a method to identify individual apnea and hypopnea events via the analysis of breathing and snoring sounds collected through a microphone, which requires no physical contact with a subject. We also explore snoring and breathing sounds over a wide frequency band, exceeding the capacity of human hearing. Hypothesis: Our primary hypothesis is that snoring sounds carry sufficient information to characterise partial and full collapses of the upper airway in OSA. We also hypothesise that snoring sounds have a frequency spectrum extending beyond the human hearing range, and the spectrum's high-frequency components can provide critical information on airway collapses. Method: We recorded streams of respiratory sounds using a bedside microphone (4 Hz–100 kHz) from patients undergoing PSG studies in a sleep laboratory. Audio streams were divided into non-overlapping segments of 5 s durations. Segments containing partial or full snores were manually extracted for further investigation. Altogether we extracted 19,715 such segments from 18 subjects. We then obtained the power spectrum of each of the segments and estimated spectral energies in 16 frequency bands covering the range: 4 Hz–35 kHz. Significantly, when the upper airway narrows during or in the vicinity of apnea/hypopnea events, spectral energy tends to shift to higher frequencies. In this research, we developed indices which capture these energy shifts and function as surrogate measures of airway patency. We then calibrated proposed sound based indices using PSG derived apnea and hypopnea events. Our indices allowed us to individually pick apnea and hypopnea events and estimate a sound based AHI which we named the frequency spectrum index. Results: Our results indicated that snoring sounds extend to frequencies beyond the upper limit of the nominal hearing range, i.e. 20 kHz (p < 0.0001). At the patient level, the frequency spectrum index shows a correlation of r = 0.80 with the PSG based hypopnea index, with the highest correlation in the frequency band of 5–15 kHz. Similarly, the frequency spectrum index showed a correlation of r = 0.82 with the PSG based obstructive apnea index, where, interestingly, the highest correlation occurs in the 25–30 kHz band, which is outside the human range of hearing band. Conclusion: The snoring sound based frequency spectrum index has the potential to characterise full and partial closures of the upper airway. Snoring sounds are spread over a wide spectrum and critical information on the full closure of the upper airway is predominantly carried in frequencies beyond human hearing capacity.

Low Level Texture Features for Snore Sound Discrimination.

Snoring is often associated with serious health risks such as obstructive sleep apnea and heart disease and may require targeted surgical interventions. In this regard, research into automatically and unobtrusively analysing the site of blockages that cause snore sounds is growing in popularity. Herein, we investigate the use of low level image texture features in classification of four specific types of snore sounds. Specifically, we explore histogram of local binary patterns (LBP) in dense grid of rectangular regions and histogram of oriented gradients (HOG) extracted from colour spectrograms for snore sound characterisation. Support vector machines with homogeneous mapping are used in the classification stage of the proposed method. Various experimental works are carried out with both LBP and HOG descriptors on the INTERSPEECH ComParE 2017 snoring sub-challenge dataset. Results presented indicate that LBP descriptors are better than the HOG descriptors in snore type detection and fusion of the LBP and HOG descriptors produces stronger results than either individual descriptor. Further, when compared to the challenge baseline and state-of-the-art deep spectrum features, our approach achieved relative percentage increases in unweighted average recall of 23.1% and 8.3% respectively.

Recurrent Neural Network for Classification of Snoring and Non-Snoring Sound Events.

Obstructive sleep apnea (OSA) is a disorder that affects up to 38% of the western population. It is characterized by repetitive episodes of partial or complete collapse of the upper airway during sleep. These episodes are almost always accompanied by loud snoring. Questionnaires such as STOP-BANG exploit snoring to screen for OSA. However, they are not quantitative and thus do not exploit its full potential. A method for automatic detection of snoring in whole-night recordings is required to enable its quantitative evaluation. In this study, we propose such a method. The centerpiece of the proposed method is a recurrent neural network for modeling of sequential data with variable length. Mel-frequency cepstral coefficients, which were extracted from snoring and non-snoring sound events, were used as inputs to the proposed network. A total of 20 subjects referred to clinical sleep recording were also recorded by a microphone that was placed 70 cm from the top end of the bed. These recordings were used to assess the performance of the proposed method. When it comes to the detection of snoring events, our results show that the proposed method has an accuracy of 95%, sensitivity of 92%, and specificity of 98%. In conclusion, our results suggest that the proposed method may improve the process of snoring detection and with that the process of OSA screening. Follow-up clinical studies are required to confirm this potential.

Obstructive Sleep Apnea Detection Using Difference in Feature and Modified Minimum Distance Classifier.

The current gold standard of Obstructive Sleep Apnea (OSA) diagnosis involves the use of a Polysomnography (PSG) system which requires the patient to stay in the hospital for overnight recording. The process is uncomfortable for the patient and it disturbs the patient's sleep pattern. On the other hand, it is well known that some acoustic features of the snoring sounds are good indicators of the presence of OSA, and a variety of acoustic OSA detection algorithms have been reported in the literature. Typically, these algorithms use multiple features and a relatively complex classifier, which are not ideal for handling the huge over-night data. In this paper, we propose an algorithm that uses a single feature and a relatively simple classifier. The proposed feature is the difference between two carefully selected Mel-frequency cepstral coefficients (MFCCs) of the snoring sound samples. The proposed classifier is derived based on a modified minimum distance criterion. The proposed algorithm has been tested with patient data. The results show that the proposed algorithm outperforms existing algorithms and is able to achieve up to 97.1% detection accuracy.

Snore Sound Analysis Can Detect the Presence of Obstructive Sleep Apnea Specific to NREM or REM Sleep.

Severities of obstructive sleep apnea (OSA) estimated both for the overall sleep duration and for the time spent in rapid eye movement (REM) and non-rapid eye movement (NREM) sleep are important in managing the disease. The objective of this study is to investigate a method by which snore sounds can be analyzed to detect the presence of OSA in NREM and REM sleep. Using bedside microphones, snoring and breathing-related sounds were acquired from 91 patients with OSA (35 females and 56 males) undergoing routine diagnostic polysomnography studies. A previously developed automated mathematical algorithm was applied to label each snore sound as belonging to either NREM or REM sleep. The snore sounds were then used to compute a set of mathematical features characteristic to OSA and to train a logistic regression model (LRM) to classify patients into an OSA or non-OSA category in each sleep state. The performance of the LRM was estimated using a leave-one-patient-out cross-validation technique within the entire dataset. We used the polysomnography-based diagnosis as our reference method. The models achieved 80% to 86% accuracy for detecting OSA in NREM sleep and 82% to 85% in REM sleep. When separate models were developed for females and males, the accuracy for detecting OSA in NREM sleep was 91% in females and 88% to 89% in males. Accuracy for detecting OSA in REM sleep was 88% to 91% in females and 89% to 91% in males. Snore sounds carry sufficient information to detect the presence of OSA during NREM and REM sleep. Because the methods used include technology that is fully automated and sensors that do not have a physical connection to the patient, it has potential for OSA screening in the home environment. The accuracy of the method can be improved by developing sex-specific models.

Snoring classified: The Munich-Passau Snore Sound Corpus

ObjectiveSnoring can be excited in different locations within the upper airways during sleep. It was hypothesised that the excitation locations are correlated with distinct acoustic characteristics of the snoring noise. To verify this hypothesis, a database of snore sounds is developed, labelled with the location of sound excitation. MethodsVideo and audio recordings taken during drug induced sleep endoscopy (DISE) examinations from three medical centres have been semi-automatically screened for snore events, which subsequently have been classified by ENT experts into four classes based on the VOTE classification. The resulting dataset containing 828 snore events from 219 subjects has been split into Train, Development, and Test sets. An SVM classifier has been trained using low level descriptors (LLDs) related to energy, spectral features, mel frequency cepstral coefficients (MFCC), formants, voicing, harmonic-to-noise ratio (HNR), spectral harmonicity, pitch, and microprosodic features. ResultsAn unweighted average recall (UAR) of 55.8% could be achieved using the full set of LLDs including formants. Best performing subset is the MFCC-related set of LLDs. A strong difference in performance could be observed between the permutations of train, development, and test partition, which may be caused by the relatively low number of subjects included in the smaller classes of the strongly unbalanced data set. ConclusionA database of snoring sounds is presented which are classified according to their sound excitation location based on objective criteria and verifiable video material. With the database, it could be demonstrated that machine classifiers can distinguish different excitation location of snoring sounds in the upper airway based on acoustic parameters.

Acoustics of snoring and automatic snore sound detection in children

Objective: Acoustic analyses of snoring sounds have been used to objectively assess snoring and applied in various clinical problems for adult patients. Such studies require highly automatized tools to analyze the sound recordings of the whole night’s sleep, in order to extract clinically relevant snore- related statistics. The existing techniques and software used for adults are not efficiently applicable to snoring sounds in children, basically because of different acoustic signal properties. In this paper, we present a broad range of acoustic characteristics of snoring sounds in children (N = 38) in comparison to adult (N = 30) patients. Approach: Acoustic characteristics of the signals were calculated, including frequency domain representations, spectrogram-based characteristics, spectral envelope analysis, formant structures and loudness of the snoring sounds. Main results: We observed significant differences in spectral features, formant structures and loudness of the snoring signals of children compared to adults that may arise from the diversity of the upper airway anatomy as the principal determinant of the snore sound generation mechanism. Furthermore, based on the specific audio features of snoring children, we proposed a novel algorithm for the automatic detection of snoring sounds from ambient acoustic data specifically in a pediatric population. The respiratory sounds were recorded using a pair of microphones and a multi-channel data acquisition system simultaneously with full-night polysomnography during sleep. Brief sound chunks of 0.5 s were classified as either belonging to a snoring event or not with a multi-layer perceptron, which was trained in a supervised fashion using stochastic gradient descent on a large hand-labeled dataset using frequency domain features. Significance: The method proposed here has been used to extract snore-related statistics that can be calculated from the detected snore episodes for the whole night’s sleep, including number of snore episodes (total snoring time), ratio of snore to whole sleep time, variation of snoring rate, regularity of snoring episodes in time and amplitude and snore loudness. These statistics will ultimately serve as a clinical tool providing information for the objective evaluation of snoring for several clinical applications.

The Finite Element Simulation of the Upper Airway of Patients with Moderate and Severe Obstructive Sleep Apnea Hypopnea Syndrome.

Objectives To investigate the snoring modes of patients with Obstructive Sleep Apnea Hypopnea Syndrome and to discover the main sources of snoring in soft tissue vibrations. Methods A three-dimensional finite element model was developed with SolidEdge to simulate the human upper airway. The inherent modal simulation was conducted to obtain the frequencies and the corresponding shapes of the soft tissue vibrations. The respiration process was simulated with the fluid-solid interaction method through ANSYS. Results The first 6 orders of modal vibration were 12 Hz, 18 Hz, 21 Hz, 22 Hz, 36 Hz, and 39 Hz. Frequencies of modes 1, 2, 4, and 5 were from tongue vibrations. Frequencies of modes 3 and 6 were from soft palate vibrations. Steady pressure distribution and air distribution lines in the upper airway were shown clearly in the fluid-solid interaction simulation results. Conclusions We were able to observe the vibrations of soft tissue and the modeled airflow by applying the finite element methods. Future studies could focus on improving the soft tissues vibration compliances by adjusting the model parameters. Additionally, more attention should be paid to vibrational components below 20 Hz when performing an acoustic analysis of human snore sounds due to the presence of these frequencies in this model.

Comparison of snoring sounds between natural and drug-induced sleep recorded using a smartphone

ObjectivesSnoring is an important clinical feature of obstructive sleep apnea (OSA), and recent studies suggest that the acoustic quality of snoring sounds is markedly different in drug-induced sleep compared with natural sleep. However, considering differences in sound recording methods and analysis parameters, further studies are required. This study explored whether acoustic analysis of drug-induced sleep is useful as a screening test that reflects the characteristics of natural sleep in snoring patients. Subjects and materialsThe snoring sounds of 30 male subjects (mean age=41.8years) were recorded using a smartphone during natural and induced sleep, with the site of vibration noted during drug-induced sleep endoscopy (DISE); then, we compared the sound intensity (dB), formant frequencies, and spectrograms of snoring sounds. ResultsRegarding the intensity of snoring sounds, there were minor differences within the retrolingual level obstruction group, but there was no significant difference between natural and induced sleep at either obstruction site. There was no significant difference in the F1 and F2 formant frequencies of snoring sounds between natural sleep and induced sleep at either obstruction site. Compared with natural sleep, induced sleep was slightly more irregular, with a stronger intensity on the spectrogram, but the spectrograms showed the same pattern at both obstruction sites. ConclusionAlthough further studies are required, the spectrograms and formant frequencies of the snoring sounds of induced sleep did not differ significantly from those of natural sleep, and may be used as a screening test that reflects the characteristics of natural sleep according to the obstruction site.

Spatial acoustic radiation of respiratory sounds for sleep evaluation.

Body posture has an effect on sleeping quality and breathing disorders and therefore it is important to be recognized for the completion of the sleep evaluation process. Since humans have a directional acoustic radiation pattern, it is hypothesized that microphone arrays can be used to recognize different body postures, which is highly practical for sleep evaluation applications that already measure respiratory sounds using distant microphones. Furthermore, body posture may have an effect on distant microphone measurement; hence, the measurement can be compensated if the body posture is correctly recognized. A spherical harmonics decomposition approach to the spatial acoustic radiation is presented, assuming an array of eight microphones in a medium-sized audiology booth. The spatial sampling and reconstruction of the radiation pattern is discussed, and a final setup for the microphone array is recommended. A case study is shown using recorded segments of snoring and breathing sounds of three human subjects in three body postures in a silent but not anechoic audiology booth.

Snore sound recognition: On wavelets and classifiers from deep nets to kernels.

In this paper, we present a comprehensive comparison of wavelet features for the classification of snore sounds. Wavelet features have proven to be efficient in our previous work; however, the benefits of wavelet transform energy (WTE) and wavelet packet transform energy (WPTE) features were not clearly established. In this study, we firstly present our updated snore sounds database, expanded from 24 patients (collected by one medical centre) to 40 patients (collected by three medical centres). We then study the effects of varying frame sizes and overlaps for extraction of the wavelet low-level descriptors, the effect of which have yet to be fully established. We also compare the performance of the WTE and WPTE features when fed into multiple classifiers, namely, Support Vector Machines (SVM), K-Nearest Neighbours, Linear Discriminant Analysis, Random Forests, Extreme Learning Machines, Kernel Extreme Learning Machines, Multilayer Perceptron, and Deep Neural Networks. Key results presented indicate that, when fed into a SVM, WTE outperforms WPTE (one-tailed z-test, p<;0.002). Further, WPTE can achieve a significant improvement when trained by a k-nearest neighbours classifier (one-tailed z-test, p <; 0.001).

Characterizing the NREM/REM sleep specific obstructive sleep apnea severity using snore sounds.

Obstructive Sleep Apnea (OSA) patients have frequent breathing obstructions and upper airway (UA) collapse during sleep. It is clinically important to estimate OSA severity separately for Rapid Eye Movement (REM) and non-REM (NREM) sleep states, but the task requires Polysomnography (PSG) which uses about 15-20 body contact sensors and subjective assessment. Almost all OSA patients snore. Vibration in narrowed UA muscles cause snoring in OSA. Moreover, as sleep states are associated with distinct breathing patterns and UA muscle tone, REM/NREM specific information must be available via snore/breathing sounds. Our previous works have shown that snoring carries significant information related to REM/NREM sleep states and OSA. We hypothesized that such information from snoring sound could be used to characterize OSA specific to REM/NREM sleep states independent of PSG. We acquired overnight audio recording from 91 patients (56 males and 35 females) undergoing PSG and labeled snore sounds as belonging to REM/NREM stages based on PSG. We then developed features to capture REM/NREM specific information and trained logistic regression (LR) classifier models to map snore features to OSA severity bands. Considering separate LR models for males and females, we achieved 94-100% sensitivity (84-89% specificity) for NREM stages at the OSA severity threshold of 30 events/h. Corresponding sensitivity for REM stages were 92-97% with specificity 83-85%. Results indicate that it is feasible to estimate severe/non-severe OSA in REM/NREM sleep based on snore/breathing sounds alone, acquired using simple bedside sound acquisition devices such as mobile phones.

Acoustic analysis of snoring sounds originating from different sources determined by drug-induced sleep endoscopy

Objective: To discuss the possibility of fundamental frequency (F0) and formant frequency (FF) to generally differentiate the sources of snoring sounds determined by drug-induced sleep endoscopy (DISE).Methods: A total of 74 snoring subjects underwent DISE and snoring sounds were recorded simultaneously. The noise-suppressed snoring sounds were analyzed and classified into different groups based on the sources of vibration identified by DISE. F0 and FFs were calculated.Results: Totally, 516 snoring sounds from three vibrating sources (the palate, combined the palate and the lateral wall, the lateral wall) of 47 patients were divided into three groups then analyzed. The levels of F0 and FFs for each group follow the order: Group 1 < Group 2 < Group 3. There was statistical difference between Group 1 and other groups in F0 and F2 (p < .05). The area under the receiver–operator curves (AUC) was F0, at 0.727, and the cut-off value was 134.2 Hz; and F2, at 0.654, and the cut-off value was 2028.0 Hz.Conclusions: F0 and the second formant frequency (F2) are found to be significantly lower in palatal snoring sound. F0 might be a significant in distinguishing palatal snoring sound from non-palatal snoring sound. F2 is more significant than F1 and F3 in identifying the sources of the snoring sounds but is less sensitive than F0.