Eur Respir J, 1996, 9, 2365–2370
Printed in UK - all rights reserved
Copyright ERS Journals Ltd 1996
European Respiratory Journal
ISSN 0903 - 1936
AAccoouussttiicc aannaallyyssiiss ooff ssnnoorriinngg ssoouunndd iinn ppaattiieennttss wwiitthh ssiimmppllee ssnnoorr--
iinngg aanndd oobbssttrruuccttiivvee sslleeeepp aappnnooeeaa
J.A. Fiz*, J. Abad*, R. Jané**, M. Riera*, M.A. Mañanas**, P. Caminal**,
D. Rodenstein+, J. Morera*
Acoustic analysis of snoring sound in patients with simple snoring and obstructive sleep
apnoea. J.A. Fiz, J. Abad, R. Jané, M. Riera, M.A. Mañanas, P. Caminal, D. Rodenstein,
J. Morera. ©ERS Journals Ltd 1996.
ABSTRACT: Snoring, a symptom which may indicate the presence of the obstruc-
tive sleep apnoea syndrome (OSA), is also common in the general population.
Recent studies have suggested that the acoustic characteristics of snoring sound
may differ between simple snorers and OSA patients. We have studied a small
number of patients with simple snoring and OSA, analysing the acoustic charac-
teristics of the snoring sound.
Seventeen male patients, 10 with OSA (apnoea/hypopnoea index (AHI) 26.2
events·h-1) and seven simple snorers (AHI 3.8 events·h-1), were studied. Full night
polysomnography was performed and the snoring sound power spectrum was ana-
Spectral analysis of snoring sound showed the existence of two different pat-
terns. The first pattern was characterized by the presence of a fundamental fre-
quency and several harmonics. The second pattern was characterized by a low
frequency peak with the sound energy scattered on a narrower band of frequen-
cies, but without clearly identified harmonics. The seven simple snorers and two
of the 10 patients with OSA (AIH 13 and 14 events·h-1, respectively) showed the
first pattern. The rest of the OSA patients showed the second pattern. The peak
frequency of snoring was significantly lower in OSA patients, with all but one OSA
patient and only one simple snorer showing a peak frequency below 150 Hz
A significant negative correlation was found between AHI and peak and mean
frequencies of the snoring power spectrum (p<0.0016 and p<0.0089, respectively).
In conclusion, this study demonstrates significant differences in the sound power
spectrum of snoring sound between subjects with simple snoring and obstructive
sleep apnoea patients.
Eur Respir J., 1996, 9, 2365–2370.
*Servei de Pneumologia, Hospital Universitario
Germans Trias i Pujol, Badalona, Spain.
**Instituto de Cibernética (UPC), Barcelona,
Spain. +Cliniques Universitaires Saint Luc,
Université Catholique de Louvain, Brussels,
Correspondence: J.A. Fiz
Servei de Pneumologia
Hospital Germans Trias i Pujol
Carretera del Canyet
s/n 08916 Badalona
Keywords: Acoustic analysis
obstructive sleep apnoea
Received: April 10 1995
Accepted after revision June 20 1996
This work was supported by grant 94/0625
In the last 15 yrs, snoring has entered the realm of
clinical medicine. Snoring may disturb the social and
family life of the snorer, who is unaware that he snores.
Snoring is a prevalent symptom, and about 50% of the
adult population snores frequently [1, 2]. One to five
percent of the adult male population suffer from the
sleep apnoea syndrome, in which snoring is a dominant
symptom. Several treatments exist for simple snoring
and for obstructive sleep apnoea (OSA), many of them
mutilating or invasive. Recent studies have suggested
that the acoustic characteristics of snoring may differ
between simple snorers and patients with OSA. Healthy
simple snorers, without apnoea episodes, showed a
power spectrum of snoring signal with a harmonic struc-
ture and a fundamental frequency that ranged 110–190
Hz . In snoring patients with OSA, frequency com-
ponents higher than 800 Hz were found .
We have studied a small number of patients with sim-
ple snoring or OSA, and analysed the acoustic charac-
teristics of the snoring sound as well as the pattern of
respiratory disturbances during sleep.
Material and methods
Seventeen lifelong nonsmoking male patients from
the sleep disorders clinic of the Germans Trias i Pujol
hospital in Badalona were studied. Ten of the subjects
were diagnosed as having OSA and seven as having
Lung function tests were performed using a spiro-
meter (PFT Horizont System) . Reference values were
those of ROCA et al. . Arterial blood gas values were
measured using a blood gas analyser (Radiometer ABL).
Full night polysomnography was performed accord-
ing to standard methods . Sleep analysis was per-
formed by visual inspection using RECHTSCHAFFEN and
KALES  paper scoring. Electroencephalogram (EEG),
electro-oculogram (EOG) and chin electromyogram (EMG)
were obtained from surface electrodes. Thoracic and
abdominal respiratory movements, as well as their sum,
were obtained using a inductive plethysmograph (Respi-
trace NIMS, Miami, USA). Calibration was performed
by means of a dry spirometer (S3371; Sensor Medics,
J.A. FIZ ET AL.
Table 1. – Anthropometric and respiratory function data
of subjects with simple snoring and patients with obstruc-
tive sleep apnoea
Simple snoring OSA
Age yrs 46±13 51±6
Height cm 171±9 166±7
Weight kg 87±22 91±21
BMI kg·m-2 29.7±7.2 32.9±7.6
FVC L 4.4±1.1 3.5±0.6
% pred 90±18 83±13
FEV1L 3.5±1.1 2.8±0.7
% pred 93±20 85±17
FEV1/FVC % 81±3 78±8
Pa,O2kPa 12.2±0.8 10.6±2.1
mmHg 91.3±5.5 79.7±16.0
Pa,CO2kPa 5.0±0.3 5.4±1.0
mmHg 37.2±2.6 40.2±7.8
Sa,O2% 96±1 94±6
pH 7.37±0.0 7.39±0.0
Values are presented as mean±SD. FVC: forced vital capacity;
BMI: body mass index; % pred: percentage of predicted value;
FEV1: forced expiratory volume in one second; Pa,O2: arterial
oxygen tension; Pa,CO2: arterial carbon dioxide tension; Sa,O2:
arterial oxygen saturation.
Table 2. – Sleep parameters of subjects with simple
snoring and patients with obstructive sleep apnoea (OSA)
Simple snoring OSA
TST min 268±66 315±68
Awake % of TST 23±13 10±8*
Stage 1 % of TST 41±18 28±17
Stage 2 % of TST 28±18 51±16
Stage 3&4 % of TST 1.3±3.3 0.3±0.9
REM % of TST 4.7±4.6 10.8±11.7
Lower Sa,O2% 92±4 80±15
Mean Sa,O2% 95±2 90±6
AHI events·h-1 3.8±1.7 26.2±20.1
of events s 26.7±16.5 42.7±27.3
of events s 17.7±3.7 20.5±3.9
Values are presented as mean±SD. TST: total sleep time;
REM: rapid eye movement (sleep); Sa,O2: arterial oxygen sat-
uration; AHI: apnoea/hypopnoea index *:p<0.05, for com-
parisons between groups (Kolmogorov-Smirnoff nonparametric
Anaheim, USA). Arterial oxygen saturation (Sa,O2) was
measured using a pulse oximeter (Oxy Shuttle; Sensor
Medics, USA) with a finger probe. Oronasal flow was
assessed using thermistors. All the signals were record-
ed on paper using a Sensormedics recorder.
In addition to the above, a miniature microphone (MKE
3010; SennHeiser, Germany) was positioned upon the
neck, 1 cm lateral to the median line at the level of the
cricoid cartilage. The microphone was enclosed in a
plastic hemisphere, 3 cm in diameter and leaving a 1 cm
distance between the microphone and the skin, avoiding
direct contact with skin surface. The hemisphere was atta-
ched to the skin using adhesive tape. The microphone
had a flat frequency response between 40 Hz and 30 kHz
(field transmission factor in open loop: 10 mv·Pa-1 ±2.5
An apnoea was defined as the absence of airflow dur-
ing ≥10 s. Apnoeas were classified as obstructive or
central according to the persistency or absence of res-
piratory movements, accompanied by a fall in Sa,O2of
≥4%. Hypopnoea was defined as a reduction in tidal
volume to ≥50% of those recorded during the preced-
ing five breaths for longer than 10 s, accompanied by
a fall in Sa,O2of ≥4%. OSA was defined as the pres-
ence of more than 10 apnoeas/hypopnoeas·h-1 of sleep,
and the total number of episodes of apnoea and hypop-
nea per hour of sleep represented the apnoea/hypop-
noea index (AHI).
The sound signal was filtered with a band-pass filter
between 10 Hz and 6 kHz (KH 39188). The sound sig-
nal together with the flow signal of the thermistor, was
stored using a videotape cassette (Racal V Store CH)
at a speed of 7.5 in·s-1. Flow and sound signals were
processed using a personal computer (PC Compaq desk
368/20e) with a sampling frequency of 12 kHz. Analysis
was limited to Stage 1 and 2 non-rapid eye movement
Snoring and flow signals were synchronized by means
of a digital clock available on videotape recorder. Three
episodes of three consecutive breaths with snoring were
randomly analysed after identification of 10 periods of
snoring. Analysis was performed using the average of
the three episodes. In patients with OSA, snoring cor-
responded to the 1st, 2nd and 3rd breath after an apnoea.
Snoring was identified by listening to the sound. Fast
Fourier Transform (FFT) was used to calculate sound
spectra of the three successive sound inspiratory episo-
des of snoring. FFT was applied on 1,024 samples.
Welch periodogram was applied using a Hanning win-
dow with an overlap of 50%.
The following parameters were measured: maximal
frequency (fmax), defined as the upper frequency con-
taining 90% of the total power of the spectrum; peak
frequency (fpeak) defined as the frequency with the max-
imum power; and mean frequency (fmean) defined as the
frequency including half the total power of the spec-
trum. Harmonics were defined as broad frequency bands
in the spectrum analysis [3, 4] with a fundamental fre-
quency. For each parameter, the variability was esti-
mated by calculating the coefficient of variation (CoV)
individually for each subject or patient. Comparisons
were performed using the nonparametric Kolmogorov-
Smirnoff test. Spearman's nonparametric correlation co-
efficient was used to verify the relationship between
sleep parameters with acoustic spectral sound parame-
ters. Results were considered significant if the p-value
was lower that 0.05.
Table 1 presents the anthropometric as well as the
lung function and blood gas values of the patients.
Patients with OSA were older, and had somewhat lower
values for forced vital capacity (FVC), forced expira-
tory volume in one second (FEV1) and arterial oxygen
tension (Pa,O2) than patients with simple snoring. How-
ever, there were no statistically significant differences
Polysomnographic data are represented in table 2.
Sleep architecture was abnormal in both groups, with
an excess of Stages 1 and 2 non-REM sleep and very
ACOUSTIC ANALYSIS OF SNORING 2367
Fig. 1. – Analysis of snoring signal from a simple snorer. a) Signal
(arbitrary units (au)) of first breath; b) power spectrum (au) of first
breath; c) signal (au) of second breath; d) power spectrum (au) of
second breath. There is a fundamental frequency and clearly identi-
Fig. 2. – Analysis of snoring signal from an obstructive sleep apnoea
(OSA) patient. a) Signal (arbitary units (au)) of first postapnoeic snore;
b) power spectrum (au) of first postapnoeic snore; c) signal (au) of
second postapnoeic snore; d) power spectrum of second postapnoeic
snore. Maximum energy is around the fundamental frequency; note
that there are no indentifiable harmonics.
00.2 0.4 0.6 0.8 11.2 1.4 1.6 1.8 2.0
Power spectrum au
100 200 300 1000
400 500 600 700 800
00.5 1.0 1.5 2.0 2.5
Power spectrum au
100 200 300 1000
400 500 600 700 800
00.2 0.4 0.6 0.8 1.0 1.2
00 100 200 300 400 500 600 700 800 900
Power spectrum au
00.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
0 100 200 300 400 500 600 700 800 900 1000
Power spectrum au
small amounts of Stages 3 and 4 non-REM sleep. REM
sleep was better preserved.
Spectral analysis of snoring showed the existence of
two different patterns. The first pattern was character-
ized by the presence of a fundamental frequency and
several harmonics through a broad frequency band (fig.
1). The second pattern was characterized by the clear
predominance of sound energy scattered on a narrower
band of frequencies, but without clearly identified har-
monics (fig. 2).
The seven simple snorers and two of the 10 patients
with sleep apnoea showed the first pattern. The two pat-
ients with the less severe OSA showed the first pattern
and had AHI of, respectively, 13 and 14 events·h-1 of
sleep. The rest of the patients with OSA showed the
second pattern of sound spectrum.
Table 3 presents the average frequency indices cal-
culated for both groups, as well as the mean of the indi-
vidual CoVs for each parameter. There was a significant
difference between patients with simple snoring and
patients with OSA in peak frequency (table 3). Using a
threshold of 150 Hz for the peak frequency, all but one
of the patients with OSA, but only one subject with sim-
ple snoring, showed values below this threshold.
All of the other indices showed a tendency for lower
values in patients with OSA, but the differences did not
reach statistical significance (table 3).
Significant negative correlations were found between
the number of AHI·h-1 of sleep and the peak and mean
frequencies of the power spectrum (fig. 3 and 4).
In this study it was found that patients with simple
snoring had a snoring sound spectrum characterized by
a fundamental frequency and the presence of harmon-
ics. Snoring analysis was performed on the first breaths
following an apnoea in patients with OSA. The pattern
decribed for simple snorers was only rarely seen in pat-
ients with OSA, who instead seemed to have a spec-
trum centred around a fundamental frequency without
harmonics. In addition, peak frequency was lower in
patients with sleep apnoea. A threshold of 150 Hz best
identifid patients with sleep apnoea from simple snor-
ers. Finally, a significant negative correlation was found
between the severity of the sleep apnoea syndrome,
assessed by the AHI, and the snoring sound spectrum.
Snoring sound is produced by oscillations in the soft
palate, pharyngeal walls and epiglottis [8–10]. Snoring
is always preceded by flow limitation . Flow limita-
tion has been attributed to the sleep-related decrease in
pharyngeal muscle tone. The pharynx may be seen as
a collapsible tube. A decrease in the tone of its walls
leads to a decrease in its cross-sectional area, with flow
limitation followed by high frequency oscillations in the
pharyngeal walls. In patients with sleep apnoea, ade-
quate airflow is compromised and is restored with the
snoring sound, only after the arousal related reopening
of the pharynx. At that time, there are partial and rapid
closures and openings of the pharyngeal lumen .
Studies on theoretical models have recently confirmed
the above hypothesis on snoring production proposed
by LIISTRO et al. .
J.A. FIZ ET AL.
Table 3. – Spectrum frequency parameters of snoring
in subjects with or without obstructive sleep apnoea syn-
Snorers CoV OSAS CoV
fpeak Hz 264±107 14 157±136 14
fmean Hz 325±58 15 223±147 16
fmax Hz 462±85 8 455±199 8
Values are presented as mean±SD for the relevant data, and
mean for the coefficient of variation. fpeak: the frequency with
the highest power; fmean: the frequency with the mean spec-
trum power; fmax: the frequency with 90% of total spectrum
power; CoV: intrasubject coefficient of variation. *: p<0.05,
Kolmogorov-Smirnoff nonparametric test.
010 20 30 40 50 60 70 80
f peak Hz
Fig. 4. – Relationship between apnoea/hypopnoea index (AHI) and
mean frequency (fmean) of spectrum in seven simple snorers and 10
obstructive sleep apnoea (OSA) patients. There is a significant nega-
tive correlation (Spearman rank order correlation: r=-0.61; p<0.0089).
Fig. 3. – Relationship between apnoea/hypopnoea index (AHI) and
peak frequency (fpeak) of spectrum in seven simple snorers and 10
obstructive sleep apnoea (OSA) patients. There is a negative rela-
tionship. There is a significant negative correlation (Spearman rank
order correlation: r=-0.70; p<0.0016).
010 20 30 40 50 60 70 80
f mean Hz
According to GAVRIELI and co-workers [12, 13], col-
lapsible tubes may show flow limitation and then high
frequency oscillations of two different types: the walls
of the tube may oscillate without complete collapse; or
there may be oscillation with a complete closure in each
cycle of oscillation. Oscillation without complete clo-
sure corresponds to the sound of a wheeze, whereas
oscillation with complete closure would correspond to
the sound of snoring. In the present study, two differ-
ent types of sound spectra were found. The first type,
characterized by a fundamental frequency and harmon-
ics, was similar to the spectrum observed during speech,
especially emission of vowels . It was characterized
by a high intensity peak corresponding to the funda-
mental frequency, followed by a series of higher fre-
quency peaks with the spectrum occupying a broad band
of frequency harmonics, (fig. 1).
The second pattern of sound spectrum was character-
ized by a low frequency peak, with the rest of the energy
of the sound spectrum scattered around a narrower range
of frequencies, without clearly identifiable harmonics
(fig. 2). This second pattern could correspond to what
has been called "relaxation oscillations" or "milking"
oscillation tubes .
In the present study, differences between peak fre-
quency of both groups (table 3) were observed. Although
this result was significant, interpatient variability was
high due to characteristics of peak frequency. Mean fre-
quency was higher in snoring patients but not statisti-
The present data show substantial differences with
respect to the results of PEREZ-PADILLA et al. . In their
study, the peak frequency in patients with simple snor-
ing was lower than in the present study, whereas the
peak in patients with OSA was higher. Moreover, in
patients with OSA, they also found a residual power at
frequencies around 1,000 Hz. Nevertheless, the results
of both studies show some similarities: simple snor-
ing was characterized in both studies by a fundamen-
tal frequency and the presence of harmonics that rarely
exceeded 500 Hz. In patients with obstructive sleep
apnoea, harmonics are much more difficult to identify.
In the present study, the second broader peak frequen-
cy about 1,000 Hz was not identified.
The main difference between the present study and
the study by PEREZ-PADILLA et al. , at least as con-
cerns snoring, rests on the method and recording equip-
ment. They placed their microphone on the manubrium
sterni, whereas we placed our microphone above the lar-
ynx. The filtering effect of the upper airway cavities is
quite different from that of the trachea and main bronchi,
necessary for the sound to reach a microphone placed
on the manubrium sterni. It is, thus, possible that the
differences in the recording equipment may be respon-
sible for the quantitative differences between the pre-
sent results and those of PEREZ-PADILLA et al. . The
difference in severity of the OSA between our patients
and theirs probably cannot explain the differences
between findings. Indeed, in the present study a sig-
nificant correlation was found between the severity of
the OSA syndrome expressed as the AHI and various
indices describing the noise spectrum. This correlation
implies that, the more severe the OSA, the lower the
frequencies describing the spectrum. Therefore, one
would expect the patients of PEREZ-PADILLA et al. .
who had a more severe OSA syndrome, to have even
lower frequencies than the patients in the present sudy.
It is known that in OSA there is oedema of the soft
palate . It can be hypothesized that, the higher the
AHI the more severe the oedema, with a corresponding
increase in the mass of the soft palate. Vibrating struc-
tures emit a sound spectrum which is related to their
mass, in such a way that, the higher the mass the lower
the frequency spectrum of the sound . This could
explain the present results both in terms of the lower
frequencies in OSA patients compared with patients with
simple snoring, and in the negative correlation between
the severity of the AHI and the sound spectrum.
Though the general spectra patterns are different in
OSA patients and simple snorers, the two patients with
the least severe form of OSA showed sound patterns
including harmonics. By contrast, only one patient with
OSA showed a peak frequency higher than 150 Hz, and
only one simple snorer showed frequencies of less than
An important caveat should be born in mind con-
cerning the present study. In patients with OSA, only
the first three breaths after a complete apnoea were
analysed. Those three breaths generally take place dur-
ing the arousal that ends each apnoea. Sleep apnoea is
precisely characterized by the occurrence of apnoeas
followed by an arousal that allows resumption of flow.
Thus, we feel it justified to have performed the present
study in this way. If future studies confirm these res-
ults, then automated acoustic analysis of snoring could
become a useful tool in clinical sleep medicine to sepa-
rate individuals with apnoea from those with simple snor-
In conclusion, we have found significant differences
in the sound spectrum of snores between patients with
simple snoring and patients with obstructive sleep apnoea.
These significant differences were observed despite the
presence of some overlap between the two groups. More-
over, there was a significant correlation between indices
describing the sound spectrum and the severity of the
obstructive sleep apnoea syndrome.
1. Lugaresi E, Cirignotta F, Coccagna C, Piana C. Some
epidemiological data on snoring and cardiorespiratory
disturbances. Sleep 1980; 3: 221–224.
2. Norton PG, Dunn EV. Snoring as a risk factor for dis-
ease: an epidemiolocial survey. Br Med J 1985; 291:
3. Dalmasso F, Benedetto G, Pogolotti R, Righini G,
Spagnolo R. Digital processing of snoring sounds. Eur
Respir J 1990; 3: 528–532.
4. Perez-Padilla JR, Slawinski E, Difrancesco LM, Feige
RR, Remmers JE, Whitelaw WA. Characteristics of
the snoring noise in patients with and without occlu-
sive sleep apnea. Am Rev Respir Dis 1993; 147: 635–644.
5. American Thoracic Society. Standardization of spiro-
metry; 1987 update. Am Rev Respir Dis 1987; 136:
6. Roca J, Sanchis J, Agusti-Vidal A, et al. Spirometric
reference from a Mediterranean population. Bull Eur
Physiopathol Respir 1986; 22: 217–224.
ACOUSTIC ANALYSIS OF SNORING 2369
7. Rechtstaffen A, Kales A (eds). A manual of standard-
ized terminology, techniques and scoring system for
sleep stages of human subjects. Bethesda, National
Institutes of Health, 1968.
8. Grotberg JB, Gavriely N. Flutter in collapsible tubes: a
theoretical model of wheezes. J Appl Physiol 1989; 66:
9. Liistro G, Stanescu DC, Veriter CL, Rodenstein D,
Aubert-Tulkens G. Characteristics of snoring in obstruc-
tive sleep apnoea patients and healthy snorers during
sleep (Abstract). Eur Respir J 1989; 2: 330.
10. Liistro, Stanescu DC, Vertier C. Pattern of simulated
snoring is different through mouth and nose. J Appl
Physiol 1991; 70: 2736–2741.
11. Liistro G, Stanescu DC, Vertier C, Rodenstein DA,
Aubert-Tulkens G. Pattern of snoring in obstructive sleep
apnea patients and in heavy snorers. Sleep 1991; 14:
12. Gavriely N, Palti Y, Alroy G, Grotberg JB. Measurement
and theory of wheezing breath sounds. J Appl Physiol:
Respirat Environ Exercise Physiol 1984; 57 (2): 481–482.
13. Gavriely N, Jensen O. Theory and measurements of
snores. J Appl Physiol 1993; 74(6): 2828–2837.
14. Fiz JA, Morera J, Abad J, et al. Acoustic analysis of
vowel emission in obstructive sleep apnea. Chest 1993;
15. Cancelli C, Pedley TJ. A seperated-flow model for col-
lapsible-tube ocillations. J Fluid Mech 1985; 157: 375–404.
16. Ryan CF, Lowe AA, Li D, Fleetham JA. Magnetic res-
onance imaging of the upper airway in obstructive sleep
apnoea before and after chronic nasal pressure therapy.
Am Rev Respir Dis 1991; 144: 939–944.
J.A. FIZ ET AL.